CN101120584B - Encoding device and method, composite device and method, and transmission system - Google Patents

Abstract

Provided are an encoding device and method, a decoding device and method, and a transmission system capable of rapidly performing an encoding and a decoding. A maximum significant digit calculating unit 63 calculates the maximum significant digits of a predetermined w quantized coefficients. A Variable Length Code (VLC) encoding unit 64 outputs a code indicating the maximum significant digits based on the calculated results. A significant digit extracting unit 65 extracts the significant digits of each of the w quantized coefficients based on the maximum significant digits. A VLC encoding unit 66 outputs a code indicating an absolute value of the w quantized coefficients based on the extracted significant digits. A sign extracting unit 67 extracts signs indicating whether each of the quantized coefficients is positive or negative from w quantized coefficients. A VLC encoding unit 68 outputs a code indicating w signs.

Description

Encoding device and method, decoding device and method, and transmission system

technical field

The present invention relates to an encoding device and method, a decoding device and method, and a transmission system, and particularly to an encoding device and method, a decoding device and method, and a transmission system in which encoding and decoding can be performed at a higher speed.

Background technique

JPEG (Joint Photographic Experts Group) 2000 is known as an encoding method in which, in the case of performing image (data) encoding, coefficients of subbands (bands) generated by performing band division processing on an input image are encoded .

In the case of encoding an image in JPEG 2000, wavelet coefficients obtained by performing wavelet transformation on an input image are quantized, and entropy encoding is further performed on the quantized coefficients obtained by the quantization.

Conventionally, for entropy coding, bit modeling called EBCOT (Embedded Block Coding with Optimized Truncation) and arithmetic coding called MQ coder are performed. That is, bit modeling is performed on the quantized coefficients, and arithmetic coding is further performed based on multiple coding paths for each bit plane. A code obtained by arithmetic coding is output as a coded image (data) (for example, Patent Document 1).

Also, in the case of decoding an image encoded by JPEG 2000, processing is performed in the reverse procedure to that of the encoding case. More specifically, entropy decoding and inverse quantization are performed on codes that are encoded image data, and wavelet inverse transform is further performed on the quantized coefficients thus obtained. The image obtained by wavelet inverse transform is output as a decoded image.

[Patent Document 1] Japanese Unexamined Patent Application Publication No. 2004-166254

Contents of the invention

The problem to be solved by the present invention

However, with the above techniques, the amount of processing at the EBCOT and MQ encoders is large, so image encoding and decoding cannot be easily performed at high speed, and, for example, when performing high-resolution HD (high-definition In the case of real-time encoding (or decoding) of images, there has been a need to provide expensive dedicated hardware.

Accordingly, an object of the present invention is to provide an encoding device and method, a decoding device and method, and a transmission system in which compression encoding and decoding of image data and output of decoded image data can be performed with less delay.

means of solving the problem

An encoding device according to a first aspect of the present invention is an encoding device for encoding second data consisting of a plurality of first data representing predetermined numerical values, the encoding device comprising: a maximum effective number of bits (maximum significant digit) output section for taking, as the maximum significant digit of the first data, the significant digit of a numerical value having the largest absolute value among the numerical values represented by each first data, and outputting a value indicating the maximum significant digit a code of digits; an absolute value output section for outputting a code indicating the absolute value of the numerical value represented by the first data; and a code output section for outputting a code indicating the sign of the numerical value represented by the first data code.

The most significant digit output section may output a code indicating whether or not the maximum significant digit has been changed as a code indicating the maximum significant digit.

In a case where the maximum significant digit has been changed, the maximum significant digit output section may output a code indicating whether the maximum significant digit has increased or decreased as the code indicating the maximum significant digit.

The maximum significant digit output section may output a code indicating a change amount of the maximum significant digit.

The maximum significant digit output section may output a code indicating a maximum significant digit of first data arranged consecutively, and output a code indicating a maximum significant digit of first data further continuously arranged from the first data.

The absolute value output section may output a code indicating a value from the lowest-order digit to the most significant digit of the numerical value represented by the first data as the code indicating the absolute value of the numerical value represented by the first data.

The absolute value output section may output codes indicating the absolute value of the numerical value represented by the first data in parallel.

The encoding method according to the first aspect of the present invention includes the steps of: taking the number of significant digits having the largest absolute value among the numerical values represented by each first data as the maximum significant number of digits of the first data, and outputting a code indicating the maximum significant number of digits; outputting a code indicating an absolute value of the numerical value represented by the first data; and outputting a code indicating a sign of the numerical value represented by the first data.

A decoding device according to a second aspect of the present invention is a decoding device for decoding second data composed of a plurality of first data representing predetermined numerical values, the decoding device including: a maximum significant digit output section for taking, as the maximum significant digit, the number of significant digits of the numerical value having the largest absolute value among the numerical values represented by each first data, decoding a code indicating the maximum significant digit of the first data, and outputting the most significant number of digits; an absolute value output section for decoding a code indicating an absolute value of a numerical value represented by the first data, and outputting an absolute value of a numerical value represented by the first data; a code output section for decoding the code indicating the sign of the numerical value represented by the first data, and outputting the data indicating the sign of the numerical value represented by the first data; and a data output section for based on the maximum significant number of digits, The absolute value of the numerical value represented by the first data and the data indicating the sign of the numerical value represented by the first data are output as the first data.

The most significant digit output section may decode, as a code indicating the maximum significant digit, a code indicating whether or not the maximum significant digit has been changed.

In the case where the maximum significant digit has changed, the maximum significant digit output section may decode a code indicating whether the maximum significant digit has increased or decreased, and a code indicating the amount of change in the maximum significant digit, as an indication of the maximum significant digit number code, and output the maximum number of significant digits based on the decoding result.

The most significant digit output section may repeatedly perform a process of decoding a code indicating a maximum significant digit of first data continuously arranged, and decoding a most significant digit of first data further continuously arranged from the first data. number code.

The absolute value output section may decode a code indicating an absolute value of a numerical value represented by the first data in parallel.

The decoding method according to the second aspect of the present invention includes the steps of: taking the number of significant digits having the largest absolute value among the numerical values represented by each first data as the maximum significant number of digits of the first data, for decoding the code indicating the maximum number of significant digits, and outputting the maximum number of significant digits; decoding the code indicating the absolute value of the numerical value represented by the first data, and outputting the absolute value of the numerical value represented by the first data ; Decode the code indicating the sign of the numerical value represented by the first data, and output data indicating the sign of the numerical value represented by the first data; and the absolute value of the numerical value represented by the first data based on the maximum significant digit value, and data indicating the sign of the numerical value represented by the first data, the first data is output.

A transmission system according to a third aspect of the present invention is a transmission system comprising: encoding means for encoding second data consisting of a plurality of first data representing predetermined numerical values and a decoding device, which is used to decode the encoded code and output second data composed of the first data; the transmission system is used to transmit the code from the encoding device to the decoding device; wherein the encoding device includes: a first a maximum significant digit output section for taking, as a maximum significant digit, a significant digit having a maximum absolute value among numerical values represented by each first data, and outputting a code indicating the maximum significant digit; the first an absolute value output section for outputting a code indicating the absolute value of the numerical value represented by the first data; and a first code output section for outputting a code indicating the sign of the numerical value represented by the first data; and Wherein the decoding means comprises: a second most significant digit output section for decoding the code which has been transmitted from the encoding means and indicating the maximum significant digit output by the first maximum significant digit output section, and outputs the maximum significant number of digits; a second absolute value output section for decoding the code that has been transmitted from the encoding device and indicating the absolute value output by the first absolute value output section, and outputs the absolute value; the second a code output section for decoding a code that has been transmitted from the encoding device and indicating a symbol output by the first code output section, and outputting a code indicating the symbol; The maximum significant number of digits output by the most significant digit output section, the absolute value of the numerical value represented by the first data output from the second absolute value output section, and the output from the second code output section and indicating the first data Data representing the sign of the numeric value, output the first data.

With the first aspect of the present invention, the number of significant digits having the largest absolute value among the numerical values represented by each first data is taken as the maximum significant number of digits of the first data, and a value indicating the maximum significant number of digits is output a code; outputting a code indicating an absolute value of the numerical value represented by the first data; and outputting a code indicating a sign of the numerical value represented by the first data.

With the second aspect of the present invention, taking the significant digit having the largest absolute value among the numerical values represented by each first data as the maximum significant digit, decoding the code indicating the maximum significant digit, and outputting the the most significant number of digits; decoding the code indicating the absolute value of the numerical value represented by the first data, and outputting the absolute value of the numerical value represented by the first data; the code indicating the sign of the numerical value represented by the first data The code is decoded, and output data indicating the sign of the numerical value represented by the first data; and based on the maximum significant digit, the absolute value of the numerical value represented by the first data, and the data indicating the sign of the numerical value represented by the first data data, output the first data.

According to the third aspect of the present invention, at the encoding device, the significant digit having the largest absolute value among the numerical values represented by each first data is taken as the maximum significant digit, and a code indicating the maximum significant digit is output ; outputting a code indicating the absolute value of the numerical value represented by the first data; and outputting a code indicating the sign of the numerical value represented by the first data; and at the decoding means, to the most significant bit transmitted from the encoding means Decode the code of the number and output the maximum significant number of digits; decode the code transmitted from the encoding device and indicate the absolute value, and output the absolute value; decode the code that has been transmitted from the encoding device and indicate the sign, and outputting data indicating the sign; and outputting first data based on the maximum significant number of digits, the absolute value, and the data indicating the sign.

Advantages of the invention

According to the first aspect of the present invention, an image can be encoded. In particular, according to the first aspect of the present invention, images can be encoded at a higher speed.

According to the second aspect of the present invention, an image can be decoded. In particular, according to the second aspect of the present invention, images can be decoded at a higher speed.

According to the third aspect of the present invention, an image can be encoded at a higher speed at the encoding device, and an image can be decoded at a higher speed at the decoding device.

Description of drawings

FIG. 1 is a block diagram showing a configuration example of an image encoding device to which the present invention is applied.

FIG. 2 is a diagram for describing subbands.

FIG. 3 is a diagram showing an example of quantization coefficients to be encoded.

Fig. 4 is a block diagram showing a configuration example of an entropy coding unit.

Fig. 5 is a flowchart for describing encoding processing.

Fig. 6 is a flowchart for describing entropy encoding processing.

FIG. 7 is a flowchart for describing w-set encoding processing.

Fig. 8 is a block diagram showing a configuration example of an image decoding device.

Fig. 9 is a block diagram showing a configuration example of an entropy decoding unit.

Fig. 10 is a block diagram showing a configuration example of a code division unit.

Fig. 11 is a block diagram showing a configuration example of a code division unit.

Fig. 12 is a flowchart for describing decoding processing.

Fig. 13 is a flowchart for describing entropy decoding processing.

Fig. 14 is a flowchart for describing w-set decoding processing.

Fig. 15 is a block diagram showing another configuration example of an entropy encoding unit.

Fig. 16 is a diagram showing an example of quantization coefficients to be encoded.

Fig. 17 is a flowchart for describing w-set encoding processing.

Fig. 18 is a flowchart for describing w-set decoding processing.

Fig. 19 is a block diagram showing the configuration of one example of an image encoding device to which the invention is applied.

Fig. 20 is an outline line diagram for schematically describing wavelet transform.

Fig. 21 is an outline line diagram for schematically describing wavelet transform.

FIG. 22 is an outline line diagram for schematically describing wavelet transform in the case of applying a lifting technique to a 5×3 filter.

FIG. 23 is an outline line diagram for schematically describing wavelet transform in the case of applying lifting technique to a 5×3 filter.

FIG. 24 is an outline line diagram for schematically describing an example in which filtering by lifting of a 5×3 filter is performed up to division level=2.

Fig. 25 is an outline line diagram for schematically describing the flow of wavelet transform and inverse wavelet transform according to the present invention.

Fig. 26 is a flowchart for describing an example of encoding processing.

Fig. 27 is a block diagram for explaining an example of an image decoding device to which the present invention is applied.

Fig. 28 is a flowchart for describing an example of decoding processing.

FIG. 29 is an outline line diagram for schematically describing parallel operations of components of an image encoding device and an image decoding device to which the present invention is applied.

Fig. 30 is a block diagram showing the configuration of an example of an image encoding device to which the present invention is applied.

FIG. 31 is an outline diagram for describing the flow of processing in the case where the rearrangement processing of wavelet coefficients is performed on the image encoding device side.

FIG. 32 is an outline diagram for describing the flow of processing in the case where the rearrangement processing of wavelet coefficients is performed on the image decoding device side.

Fig. 33 is a block diagram showing the configuration of an example of an image encoding device to which the present invention is applied.

Fig. 34 is a block diagram showing the configuration of an example of an image decoding device to which the present invention is applied.

Fig. 35 is a diagram for describing an example of how encoded data is exchanged.

FIG. 36 is a diagram showing a configuration example of a packet.

FIG. 37 is a block diagram showing the configuration of an example of a digital triax system to which the present invention is applied.

Fig. 38 is a block diagram showing the configuration of an example of a wireless transmission system to which the present invention is applied.

FIG. 39 is a diagram showing the configuration of an example of a home game machine to which the present invention is applied.

Fig. 40 is a diagram showing a configuration example of an information processing system to which the present invention is applied.

reference sign

11 image coding device

23 entropy coding units

Line 61 determines the unit

62 VLC coding units

63 maximum significant digit calculation unit

64 VLC coding units

65 effective digit extraction unit

66 VLC coding units

67 symbol extraction units

68 VLC coding units

111 image decoding device

151 code segmentation unit

152 lines to determine the unit

154VLC decoding unit

155VLC decoding unit

156VLC decoding unit

301 buffer

401 image coding device

410 wavelet transform unit

411 midway calculation buffer unit

412 Coefficient Rearrangement Buffer Units

413 Coefficient Rearrangement Unit

414 rate control unit

415 entropy coding units

420 image decoding device

421 entropy decoding unit

422 Coefficient Buffer Units

423 wavelet inverse transform unit

430 image coding device

431 Coded Rearrangement Buffer Unit

432 Code Rearrangement Unit

500 transfer units

501 triaxial cable

502 Camera Control Unit

510 video signal encoding unit

511 video signal decoding unit

526 video signal decoding unit

527 video signal encoding units

600 transfer units

601 receiving device

602 video signal coding unit

612 wireless module unit

621 wireless module unit

624 video signal decoding unit

700 camera installation

701 home game console main unit

Detailed ways

FIG. 1 is a block diagram showing the configuration of an example of an image encoding device to which the present invention is applied.

The image encoding device 11 has a wavelet transform unit 21 , a quantization unit 22 , and an entropy encoding unit 23 .

For example, an image (data) serving as a component signal that has been subjected to DC level shifting as necessary is input to the wavelet transform unit 21 . The wavelet transformation unit 21 performs wavelet transformation on an input image, and divides the image into a plurality of subbands. The wavelet transform unit 21 supplies the wavelet coefficients of the subbands obtained by the wavelet transform to the quantization unit 22 .

The quantization unit 22 quantizes the wavelet coefficient supplied from the wavelet transform unit 21 , and supplies the quantized coefficient obtained as a result thereof to the entropy encoding unit 23 .

The entropy encoding unit 23 entropy-encodes the quantization coefficient supplied from the quantization unit 22, and outputs the encoding thus obtained as an encoded image (data). After rate control processing, the images output from the entropy encoding unit 23 may be packetized and recorded, or supplied to other devices (not shown) connected to the image encoding device 11, for example.

Next, entropy encoding performed by the entropy encoding unit 23 in FIG. 1 will be described with reference to FIGS. 2 and 3 .

For example, as shown in FIG. 2 , one subband is configured by six lines, lines L1 to L6 , and the positions of pixels corresponding to the lines in the xy coordinate system are taken as (x, y). Now, in the plot of each row, take the x-coordinate of the left end position as 0, and take the y-coordinate of row L1 as 0.

Quantization coefficients in the bit-plane representation and at each position (x, y) of the subband are input from the quantization unit 22 to the entropy encoding unit 23 in raster scan order from line L1 to line L6.

In other words, first, the quantization coefficient corresponding to the left end coordinate (0, 0) of the line L1 is input to the entropy encoding unit 23 . Next, the quantization coefficient corresponding to the coordinate (1, 0) adjacent to the right side of the coordinate (0, 0) is input to the entropy encoding unit 23, and corresponds to the coordinate adjacent to the coordinate in which the quantization coefficient has been input. The quantized coefficients are sequentially input to the entropy encoding unit 23 up to the coordinates of the right end of the line L1. When all the quantization coefficients of the coordinates on the line L1 are input, sequentially from the coordinates on the left end of the line L2 (0, 1) to the coordinates on the right end, the quantization coefficients corresponding to each coordinate on the line L2 are input to the entropy encoding unit 23 , and similarly, from line L3 to line L6 , quantization coefficients corresponding to coordinates on each line are input to the entropy encoding unit 23 .

For example, in FIG. 3, as shown on the upper left of the figure, when twelve quantized coefficients are input to the entropy encoding unit 23 in order from the coordinates on the left end of line L1 in FIG. 2, entropy encoding The unit 23 encodes the quantized coefficients in increments of a predetermined number w (w=4 in FIG. 3 ) determined in advance.

Now, the quantized coefficients illustrated on the upper left in FIG. 3 are each expressed in such a manner that the absolute value of its code is divided into binary digits (bit-plane expression), and by the example shown in FIG. 3, one row (in FIG. 2 The quantization coefficients "-0101", "+0011", "-0110", "+0010", "+0011", "+0110", "0000", "-0011", "+1101" of the row L1) , “-0100”, “+0111”, and “-1010” are input into the entropy encoding unit 23 in sequence.

Each quantization coefficient is composed of a quantization coefficient code expressed as "+" (positive) or "-" (negative) (hereinafter referred to as a sign (Sign) of the quantization coefficient), and an absolute value of the quantization coefficient expressed in binary. In FIG. 3 , among the bits representing the value of each digit of the absolute value of the quantization coefficient, the uppermost bit in the figure represents the highest-order bit (bit of the highest-order digit). Therefore, for the quantization coefficient "-0101", its sign is "-", and the absolute value expressed in binary is "0101", so the decimal representation of the quantization coefficient is "-5".

First, the entropy encoding unit 23 determines whether (the absolute value of) the quantization coefficients of an input line are all 0, and outputs a code indicating whether or not all the quantization coefficients of the line to be encoded are 0 based on the determination result. In a case where it is determined that the quantization coefficients are all 0, the entropy encoding unit 23 outputs 0 as a code indicating whether all the quantization coefficients of the row are 0, and ends the encoding of the quantization coefficients of the row currently being performed. And, in a case where it is determined that not all quantized coefficients have values of 0 (not only quantized coefficients of 0 exist), the entropy encoding unit 23 outputs 1 as a code indicating whether all quantized coefficients of the row are 0.

In the case where these twelve quantized coefficients shown on the upper left in the figure are input, the variable delay of the quantized coefficients of the input row is not only 0, so the entropy encoding unit 23 outputs 1 as a code, as shown on the upper right in the figure of.

When the code 1 indicating that the quantization coefficients are not all 0 is output as the code indicating whether all the quantization coefficients of the row are 0, then, the entropy encoding unit 23 performs the first four (w) input quantization coefficients "- 0101", "+0011", "-0110", and "+0010" to perform encoding.

The entropy coding unit 23 compares the most significant digits (the value of the variable B in FIG. 3 ) of the four consecutive quantization coefficients input this time with the most significant digits of the four (w) quantization coefficients of the previous (input) encoding number, determine whether the maximum significant digit has changed, and output a code indicating the maximum significant digit of the quantization coefficient.

Now, the maximum significant number of bits is the maximum significant number of bits of the quantization coefficient having the largest value among four (w) quantization coefficients to be commonly encoded. In other words, the maximum significant digit indicates at which digit the highest-order 1 is located for the quantization coefficient having the largest absolute value among the four quantization coefficients. Thus, the maximum significant number of digits of the four quantization coefficients "-0101", "+0011", "-0110", and "+0010" to be commonly encoded is, for example, "3", which is for the Quantization coefficient "-0110", the digit where the highest order 1 is located.

And, the code indicating the maximum significant digit of the quantization coefficient is composed of a code indicating whether the maximum significant digit has been changed, a code indicating whether the maximum significant digit has been increased or decreased, and a code indicating the maximum significant digit The code of the change amount, and in the case that the maximum significant digit has not changed, the code indicating whether the maximum significant digit has increased or decreased and the code indicating the change amount of the maximum significant digit are not output.

From the comparison result of the maximum significant digits, the entropy encoding unit 23 outputs code 1 indicating that the maximum significant digits have been changed when the maximum significant digits have been changed, and outputs Code 0 indicating that the maximum significant digit has not changed.

And, regarding the determination of whether the maximum significant number of digits has been changed, in the case where four quantization coefficients are input for the first time, that is, in the case of inputting the first quantization coefficient of a subband to be encoded (for example, in In the case where four quantization coefficients are input from the left in the order of line L1 in FIG. is set to 0.

Therefore, the entropy encoding unit 23 compares the maximum significant number of digits 3 of the four quantization coefficients "-0101", "+0011", "-0110", and "+0010" input this time with the four quantization coefficients of the previous encoding. The maximum significant digit is 0, and the output code is 1, because the maximum significant digit has changed.

And, the entropy encoding unit 23 outputs a code indicating whether the maximum significant digit number has increased or decreased after the code 1 indicating that the maximum significant digit number has changed. Here, entropy encoding section 23 outputs 0 when the maximum significant number of bits increases, and outputs 1 when the maximum significant number of bits decreases.

The previous maximum significant digit is 0, and the current maximum significant digit is 3, so as shown in the upper right of the figure, the entropy encoding unit 23 outputs code 0, which indicates that the maximum significant digit increases.

Furthermore, when outputting the code indicating whether the maximum significant digit has increased or decreased, the entropy encoding unit 23 outputs a code indicating how much the maximum significant digit has increased or decreased, that is, a code indicating the change amount of the maximum significant digit. Specifically, for the change amount (ie increase or decrease) of the maximum significant digit as n, the entropy encoding unit 23 outputs (n−1) code 0s, and outputs code 1 after these 0s.

In the case of encoding the first four quantized coefficients shown in FIG. 3, the change amount of the maximum significant number of bits is 3 (=3-0), so the entropy encoding unit 23 outputs two (=3-1) 0, and then output 1 as the code.

Next, the entropy encoding unit 23 outputs a code indicating the maximum number of significant digits of the absolute value of each of the four (w) quantized coefficients to be encoded this time. That is, the entropy encoding unit 23 outputs, with respect to each quantized coefficient, a number indicating each digit of the absolute value of the quantized coefficient from the largest digit to the smallest digit in order of the significant digits indicated by the largest significant digit. code for the value.

The quantization coefficients to be encoded this time are "-0101", "+0011", "-0110", and "+0010", so the entropy encoding unit 23 first outputs The code for the largest significant digit of an absolute value. Here, the maximum significant digit is 3, so the entropy encoding unit 23 outputs the value "1" of the largest digit (ie, the third number) among the significant digits indicated by the maximum significant digit of the quantization coefficient "-0101" , the value "0" of the digit below the largest digit (the second digit), and the value "1" of the lowest order digit. Thus, the code "101" indicating the number of significant digits of the absolute value of the quantization coefficient "-0101" is output.

In the same manner, the entropy encoding unit 23 sequentially outputs the codes "011", "110", and "010" for the significant number of bits indicating the absolute values of the quantization coefficients "+0011", "-0110", and "+0010". . Therefore, "101011110010" is output as a code indicating the maximum number of significant digits of the absolute value of each of "-0101", "+0011", "-0110", and "+0010". Thus, the entropy encoding unit 23 outputs a code having a length corresponding to the maximum significant digits of the four quantization coefficients to be encoded, as a code indicating the absolute value of the quantization coefficient.

Finally, the entropy encoding unit 23 outputs a code indicating the sign of each of the four (w) quantized coefficients whose absolute value is not 0. Now, the entropy encoding unit 23 outputs this code 0 when the sign of the quantized coefficient is "+" (positive), and outputs this code 1 when the sign of the quantized coefficient is "-" (negative).

The quantized coefficients to be encoded this time are "-0101", "+0011", "-0110", and "+0010", and the signs of these quantized coefficients are negative, positive, negative, and positive in turn, so as shown in the figure As shown in the upper right of , the entropy encoding unit 23 outputs "1010" as a code indicating the sign of each quantized coefficient.

When the four quantized coefficients input for the first time are encoded, the entropy encoding unit 23 then encodes the next consecutive four quantized coefficients "+0011", "+0110", "0000", and "-0011".

In the same manner as in the case of encoding the quantization coefficients input for the first time (previous time), the entropy encoding unit 23 first compares the maximum significant digits of the four (w) quantization coefficients newly input this time with the four quantization coefficients of the previous encoding. (w) The maximum number of significant digits for quantization coefficients.

The maximum significant number of digits of the four (w) quantization coefficients "+0011", "+0110", "0000", and "-0011" input this time is "3", which is for quantization with the largest absolute value The coefficient "+0110", the digit at which the highest-order 1 is located, and this is the same as the maximum significant digit of the quantization coefficient of the previous encoding, so the entropy encoding unit 23 outputs code 0, which indicates that the maximum significant digit has not changed.

Next, the entropy encoding unit 23 outputs the code "011110000011", which indicates each of the four (w) quantized coefficients "+0011", "+0110", "0000", and "-0011" to be encoded this time The codes "011", "110", "000", and "011" of the most significant digit of an absolute value are arranged in order.

When the code indicating the absolute value of the quantization coefficient has been output, the entropy encoding unit 23 outputs the code indicating the sign of each of the four (w) quantization coefficients whose absolute value is not 0.

The quantization coefficients to be encoded this time are "+0011", "+0110", "0000", and "-0011", and the absolute value of the third quantization coefficient "0000" is 0, so the entropy encoding unit 23 outputs Code "001", which indicates the signs (positive, positive, negative) of these quantization coefficients "+0011", "+0110", and "-0011" which are not 0.

When the four quantized coefficients "+0011", "+0110", "0000", and "-0011" are encoded, the entropy encoding unit 23 then encodes the subsequent four consecutive quantized coefficients "+1101", "-0100 ", "+0111", and "-0101" to encode.

The entropy coding unit 23 first compares the maximum significant digits of the four (w) quantization coefficients newly input this time with the maximum significant digits of the four quantization coefficients encoded last time.

The maximum significant number of digits of the four (w) quantization coefficients "+1101", "-0100", "+0111", and "-0101" input this time is "4", which is for the The quantization coefficient "+1101", the digit at which the highest-order 1 is located, and this is different from the maximum significant digit "3" of the quantization coefficient encoded last time, so the entropy encoding unit 23 outputs code 1 indicating the maximum significant digit changed.

And, the previous maximum significant digit was 3, and this time the maximum significant digit is 4, so the entropy encoding unit 23 outputs code 0, indicating that the maximum significant digit is increased, as shown on the right side of the figure.

Furthermore, the entropy encoding unit 23 outputs a code indicating how much the maximum significant number of bits has increased. In this case, the change amount of the most significant digit is 1 (=4-3), so the entropy encoding unit outputs 0 (=1-1) zeros, and also outputs 1 (ie, outputs code 1).

Next, the entropy encoding unit 23 outputs the code "1101010001111010", which indicates the number of the four (w) quantization coefficients "+1101", "-0100", "+0111", and "-1010" to be encoded this time. The codes "1101", "0100", "0111", and "1010" of the most significant digits of each absolute value are arranged in order.

When the code indicating the absolute value of the quantization coefficient is output, the entropy encoding unit 23 outputs the code indicating the sign of each of the four (w) quantization coefficients whose quantization coefficient is not 0.

The quantized coefficients to be coded this time are "+1101", "-0100", "+0111", and "-1010", and the signs of these quantized coefficients are positive, negative, positive, and negative in turn, so as shown in the figure As shown on the lower right, the entropy encoding unit 23 outputs "0101" as a code indicating the sign of each quantization coefficient.

The entropy encoding unit 23 thus encodes the continuous predetermined number (w) of input quantized coefficients. Thus, when a code indicating whether all quantization coefficients of a row to be encoded are 0 is output from the entropy encoding unit 23, and a code indicating that not all quantization coefficients of the row are 0 is output, next, a code indicating w quantization coefficients is output A code of the maximum significant number of digits, a code indicating the absolute value (bit-plane expression) of w quantization coefficients, and a code indicating the sign of the quantization coefficients.

For the codes indicating the maximum significant digits of these w quantization coefficients, the codes indicating the absolute values of the w quantization coefficients, and the codes indicating the signs of the quantization coefficients, the maximum significant digits indicating the next w quantization coefficients are repeatedly output A code indicating the absolute value of the w quantized coefficients, and a code indicating the sign of the quantized coefficients until all the quantized coefficients of the row are coded.

Now, it has been described that quantization coefficients are encoded in raster scan order, but the order in which quantization coefficients are encoded is not necessarily raster scan order. For example, in the case where the quantized coefficients of the subbands shown in FIG. 2 are coded, an arrangement may be made in which positions (0, 0), (0, 1), (0, 2), and (0 , 3) (i.e., at the position at the left end of the figure for each of the lines L1 to L4) are coded, and then, at positions (1,0), (1,1), (1,2), The quantized coefficients of and (1, 3) are encoded, and so on, the quantized coefficients at four positions vertically aligned in the figure are adopted as w quantized coefficients, and are sequentially encoded w one at a time.

The entropy encoding unit 23 shown in FIG. 1 that performs processing such as described above is configured as shown in FIG. 4 in more detail.

The entropy coding unit 23 includes a row determination unit 61, a VLC (Variable Length Code) coding unit 62, a maximum significant digit calculation unit 63, a VLC coding unit 64, a significant digit extraction unit 65, a VLC coding unit 66, a symbol extraction unit 67, VLC encoding unit 68, and code linking unit 69.

Quantization coefficients output from the quantization unit 22 ( FIG. 1 ) are supplied (input) to the row determination unit 61 , maximum significant digit calculation unit 63 , significant digit extraction unit 65 , and sign extraction unit 67 .

The line determination unit 61 determines whether or not all quantization coefficients of a line to be encoded input from the quantization unit 22 are 0, and supplies information indicating the determination result to the VLC encoding unit 62 .

Based on the information indicating the result of the judgment made by the row determination unit 61 , the VLC encoding unit 62 outputs a code indicating whether all quantization coefficients of the row to be encoded are 0 to the code linking unit 69 .

The maximum significant digit calculation unit 63 calculates the maximum significant digit number of w consecutive quantization coefficients input from the quantization unit 22 , and supplies information indicating the calculation result thereof to the VLC encoding unit 64 and the significant digit extraction unit 65 .

Based on the information indicating the calculation result from the maximum significant digit calculation unit 63 , the VLC encoding unit 64 supplies codes indicating the maximum significant digits of the w quantization coefficients to the code linking unit 69 .

Based on the information indicating the calculation result from the maximum significant digit calculation unit 63, the significant digit extraction unit 65 extracts the significant digits of the w quantization coefficients supplied from the quantization unit 22, and (data of) the significant digits of the extracted quantization coefficients It is supplied to the VLC encoding unit 66 and the symbol extracting unit 67 .

The VLC encoding unit 66 encodes the absolute values of the quantized coefficients based on the significant digits of these quantized coefficients from the significant-digit extracting unit 65 , and supplies the absolute values of the quantized coefficients thus obtained to the code linking unit 69 .

The sign extraction unit 67 extracts the sign of the quantization coefficient supplied from the quantization unit 22 based on the significand of the quantization coefficient from the significand extraction unit 65 , and supplies (the data of) the extracted sign to the VLC encoding unit 68 .

The VLC encoding unit 68 encodes (the data of) the extracted symbol supplied from the symbol extracting unit 67 , and supplies the code indicating the sign of the quantized coefficient thus obtained to the code linking unit 69 .

The code linking unit 69 links a code indicating whether or not all quantization coefficients of the row are 0, a code indicating the maximum number of significant digits supplied from the VLC encoding unit 62, VLC encoding unit 64, VLC encoding unit 66, and VLC encoding unit 68, respectively. , a code indicating the absolute value of the quantization coefficient, and a code indicating the sign of the quantization coefficient, and output them as encoded image (data).

Next, encoding processing by the image encoding device 11 ( FIG. 1 ) will be described with reference to the flowchart in FIG. 5 . This encoding process starts when an image (data) to be encoded is input to the wavelet transform unit 21 .

In step S11 , the wavelet transformation unit 21 subjects the input image to wavelet transformation, divides the input image into a plurality of subbands, and supplies the wavelet coefficient of each subband to the quantization unit 22 .

In step S12 , the quantization unit 22 quantizes the wavelet coefficients supplied from the wavelet transform unit 21 , and supplies the quantized coefficients obtained as a result thereof to the entropy encoding unit 23 . Thus, for example, quantization coefficients at each position in a subband in the bit-plane expression that has been described with reference to FIG. 3 are input to the entropy encoding unit 23 .

In step S13, the entropy encoding unit 23 performs entropy encoding processing, and ends the encoding processing. Although the details of the entropy encoding process will be described later, in the entropy encoding process described with reference to FIG. A code indicating whether all quantization coefficients of the line to be encoded are 0, a code indicating the maximum significant number of digits of w quantization coefficients, a code indicating the absolute value of the quantization coefficient, and an indication are output as encoded image (data). Code for the sign of the quantization coefficient.

Thus, the image encoding device 11 encodes and outputs the input image.

Next, entropy encoding processing corresponding to the processing in step S13 in FIG. 5 will be described with reference to FIG. 6 .

In step S12 in FIG. 5, the quantization coefficient output from the quantization unit 22 is supplied (input) to the row determination unit 61, the maximum significant digit calculation unit 63, and the significant digit extraction unit 65 of the entropy encoding unit 23 (FIG. 4). , and a symbol extraction unit 67.

In step S41, the row determination unit 61 takes the variable y indicating the row of the subband to be encoded now as y=0, and stores it.

For example, in the case of encoding the quantized coefficients of the subbands shown in FIG. 2 , the row determination unit 61 takes the variable y indicating these rows (row L1 to row L6 ) as y=0. Note that the row y indicated by the variable y indicates the row whose y-coordinate is y at each position (x, y) on the row of the subband. Therefore, when the variable y stored in the row specifying unit 61 is, for example, y=0, the row indicated by the variable becomes the row L1 whose y-coordinate is 0 at each position on the row.

In step S42, the maximum significant digit calculation unit 63 fetches the maximum significant digits of the first input w quantization coefficients on the row (y−1) preceding the row y indicated by the variable y stored in the row determination unit 61 The variable Binit of is stored as Binit=0.

For example, in the case where row (y-1) is L1 shown in FIG. 2 , the value of the variable Binit indicating the maximum number of significant digits of w quantization coefficients input first on row (y-1) is obtained from The w quantized coefficients at the left edge of line L1 in the figure, that is, the maximum significant digits of the w quantized coefficients at positions (0, 0), (1, 0), ..., (w-1, 0) . And, when the variable y stored in the row determination unit 61 is y=0, the row (y−1) does not exist, so the value of the variable Binit is Binit=0.

In step S43 , the row determination unit 61 determines whether or not the (absolute value of) quantization coefficients of the row y indicated by the variable y stored therein are all 0. For example, in a case where row y is row L1 shown in FIG. 2 , row determination unit 61 determines all quantization coefficients to be 0 if all quantization coefficients at position (x, y) on row L1 are 0.

In the case where it is determined in step S43 that the quantization coefficients are all 0, the row determination unit 61 generates information meaning that all the quantization coefficients are 0, and supplies it to the VLC encoding unit 62 and the maximum significant digit calculation unit 63, and the flow Proceed to step S44.

In step S44, based on the information from the row determination unit 61 meaning that all the quantization coefficients are all 0, the VLC encoding unit 62 outputs (supplies) to the code 0 indicating that all the quantization coefficients of the row to be encoded are 0. Link unit 69. The code linking unit 69 takes the code 0 supplied from the VLC encoding unit 62 and outputs it without change, serving as a code obtained as a result of encoding the quantized coefficients of the row y.

In step S45, based on information from row determination unit 61 meaning that all quantization coefficients are 0, maximum significant number of digits calculation unit 63 sets the value of variable Binit stored therein to Binit=0, and updates the variable Binit .

In step S46, the line determination unit 61 determines whether there are any unprocessed lines among the lines of the encoded subband. That is, the line determination unit 61 determines whether quantization coefficients of all lines of the subband to be encoded have been encoded. For example, in the case of encoding the quantized coefficients of subbands shown in FIG. line.

In a case where it is determined in step S46 that there is an unprocessed line, the line determination unit 61 advances the flow to step S47 to encode the quantization coefficient at each position on the next line, that is, line (y+1).

In step S47, the row determination unit 61 increments the variable y indicating the stored row by y=y+1, returns the flow to step S43, and executes the above-described subsequent processing again.

On the contrary, in the case where it is determined in step S46 that there are no unprocessed lines, the quantized coefficients of all the lines constituting the subband have been encoded, and therefore the line determination unit 61 ends this entropy encoding process, and the flow returns to the entropy encoding process in FIG. Step S13, and the encoding process ends.

And, when it is determined in step S43 of FIG. 6 that the quantized coefficients of row y are not all 0 (there are non-zero quantized coefficients), the row determination unit 61 generates a meaning that the quantized coefficients are not all 0 (there are non-zero quantized coefficients). The information of the quantization coefficient) is supplied to the VLC encoding unit 62 and the maximum significant number of digits calculating unit 63, and the flow is advanced to step S48.

In step S48, based on the information from the line determination unit 61 that means that the quantization coefficients are not all 0, the VLC encoding unit 62 outputs (supplies) to the code 1 indicating that the quantization coefficients of the line to be encoded are not all 0. Link unit 69.

In step S49, based on the information from row determination unit 61 that means that all quantization coefficients are 0, maximum significant number of digits calculation unit 63 sets the value of variable x to x=0, and stores this variable x, where The variable x indicates the x coordinate of the position on row y of the first quantization coefficient to be input among the w quantization coefficients to be coded now.

For example, in the case where the line y is the line L1 shown in FIG. 2, the value of the variable x stored by the maximum significant digit calculation unit 63 indicates w consecutive positions (x, 0) on the line L1 to be encoded. ), (x+1, 0), .

And, in step S49, the maximum significant digit calculation unit 63 sets the value of the variable B to B=Binit, and stores the variable B indicating the maximum significant digit of the w quantization coefficients encoded last time. That is, the maximum significant digit calculation unit 63 updates the value of the variable B as the value of the variable Binit stored therein, and stores the updated value of the variable B.

When updating the value of variable B, maximum significant digit calculation unit 63 supplies information indicating the updated value of variable B (maximum significant digit) to VLC encoding unit 64 and significant digit extraction unit 65 . Also, the VLC encoding unit 64 and the significant digit extraction unit 65 each store the value of the variable B supplied from the maximum significant digit calculation unit 63 .

In step S50, the entropy encoding unit 23 performs w-set encoding processing. Although details of the w-set encoding process will be described later, in the w-set encoding process, the entropy encoding unit 23 encodes w consecutive quantization coefficients on the row y indicated by the variable y stored in the row determining unit 61 .

Now, by taking the position on the line y identified by the variable y stored in the line determination unit 61 and the variable x stored in the maximum significant digit calculation unit 63 as (x, y), w consecutive positions of the line y are consecutive positions (x, y), (x+1, y), ..., (x+w-1, y) on row y. That is to say, in the w-set encoding process, the entropy encoding unit 23 pairs each of the positions (x, y), (x+1, y), ..., (x+w-1, y) The quantization coefficient at is encoded.

In step S51, the maximum significant number of digits calculation unit 63 determines whether or not there is an unprocessed quantization coefficient on line y. That is, the maximum significant number of digits calculation unit 63 determines whether or not all quantization coefficients stored at positions on the row y indicated by the variable y in the row determination unit 61 have been encoded.

In a case where it is determined in step S51 that there are unprocessed quantized coefficients on the line y, the maximum significant number of digits calculation unit 63 advances the flow to step S52 to encode the next w quantized coefficients.

In step S52, the maximum significant digit calculation unit 63 takes the variable x stored therein as x=x+w, and returns the flow to step S50. Therefore, the quantized coefficients at each of the positions (x+w, y), (x+w+1, y), ..., (x+2w-1, y) on the line y are determined in the following step S50 is encoded in the processing.

And, in the case where it is determined in step S51 that there is no unprocessed quantized coefficient on line y, the quantized coefficients at all positions on line y have already been encoded, so the maximum significant number of digits calculation unit 63 returns the flow to step S46, and perform subsequent processing.

Thus, the entropy encoding unit 23 encodes the quantized coefficients at each position of the subband in raster scan order every predetermined number.

Thus, by encoding the quantization coefficients at each position of the subband in raster scan order every predetermined number, the input quantization coefficients can be processed in the order in which they are input, and the delay due to the encoding of the quantization coefficients can be reduced .

Next, w-set encoding processing corresponding to the processing of step S50 in FIG. 6 will be described with reference to the flowchart shown in FIG. 7 .

In step S81, the maximum significant digit calculation unit 63 takes the position on the row y identified by the variable x stored therein as (x, y), and at w consecutive positions (x, y), (x+1, y), ..., (x+w-1, y), the effective number of digits of the quantized coefficient with the largest absolute value is taken as the value of the variable Bnew, and this variable Bnew is stored, wherein the variable Bnew indicates that it will be encoded now The maximum number of significant digits of the w quantization coefficients.

And, the maximum significant digit calculation unit 63 supplies the obtained maximum significant digit number of w quantization coefficients, that is, the value of the variable Bnew to the VLC encoding unit 64 and the significant digit extraction unit 65 .

For example, in the case where quantization coefficients at w consecutive positions are quantization coefficients "-0101", "+0011", "-0110", and "+0010" among these quantization coefficients as shown in FIG. The quantization coefficient of the absolute value is "-0110", and its significant digit is "3", which is the digit at which the highest-order 1 is located for "-0110", so the value of the variable Bnew is set to 3.

In step S82, the VLC encoding unit 64 determines whether B=Bnew. That is, the VLC encoding unit 64 determines whether the value of the variable B indicating the maximum significant digits of the w quantization coefficients encoded last time is the same as the w quantization coefficients to be encoded now supplied from the maximum significant digit calculation unit 63 The value of variable Bnew with the most significant digit is the same.

In the case where it is determined in step S82 that B=Bnew, the VLC encoding unit 64 advances the flow to step S83, and outputs to the code linking unit 69 a code 0 indicating the maximum number of significant digits of the w quantization coefficients to be encoded from now on . Upon outputting code 0 indicating the maximum significant number of digits, the VLC encoding unit 64 skips the processing in step S84 to the processing in step S88, and advances the flow to step S89.

On the contrary, in the case where it is determined that B=Bnew does not hold in step S82, the VLC encoding unit 64 advances the flow to step S84, and (since the maximum significant number of digits has changed) outputs code 1 to the code linking unit 69 indicating the maximum Significant digits have changed.

In step S85, the VLC encoding unit 64 obtains integers n and m, which satisfy the following expression (1).

Bnew＝B+(n+1)×(-1)^m ...(1)

Here, the symbol "^" in Expression (1) represents an exponent. Thus, (-1)^m means (-1) raised to the mth power.

For example, in the case of Bnew=3 and B=0, n=2 and m=0 are obtained as n and m satisfying expression (1). Comparing the variable Bnew and the variable B, the greater the difference between the absolute value of the value of the variable Bnew and the absolute value of the value of the variable B, the greater the value in the expression (1). It can thus be said that the value of n indicates the amount of change in the maximum number of significant digits. And, when the value of the variable Bnew is larger than the value of the variable B, the value of m is 0, and conversely, when the value of the variable Bnew is smaller than the value of the variable B, the value of m is 1. It can therefore be said that the value of m in Expression (1) indicates whether the maximum significant number of digits has been increased or decreased.

In step S86, the VLC encoding unit 64 outputs the value of m satisfying Expression (1) as a 1-bit code to the code linking unit 69 as a code indicating whether the maximum significant number of digits is increased or decreased. For example, in the case where the value of m satisfying expression (1) is 0, the VLC encoding unit 64 outputs code 0, which indicates that the maximum significant number of bits has increased.

In step S87, the VLC encoding unit 64 outputs to the code linking unit 69 as many zeros as the value of n satisfying expression (1), followed by one 1, as a code indicating the amount of change in the maximum significant number of digits. That is, the VLC encoding unit 64 outputs n 0s and one 1 as a code indicating the amount of change in the maximum significant number of digits.

For example, in the case where the value of n satisfying expression (1) is 2, the VLC encoding unit 64 outputs “001” to the code linking unit 69 as a code indicating the amount of change in the maximum significant number of digits.

Thus, the VLC encoding unit 64 outputs a code indicating that the maximum significant digit has been changed, a code indicating whether the maximum significant digit has been increased or decreased, and a code indicating the change amount of the maximum significant digit to the code linking unit 69 as instructions The code of the most significant number of bits of the w quantization coefficients to be coded from now on.

In step S88, the maximum significant digit calculation unit 63 sets the stored value of the variable B to B=Bnew, and advances the flow to step S89. That is, the maximum significant digit calculation unit 63 updates the stored value of the variable B to the value of the variable Bnew stored therein. Furthermore, VLC encoding section 64 and significant digit extracting section 65 also set the stored value of variable B to B=Bnew.

When the value of the variable B is set to B=Bnew in step S88, or when a code indicating the maximum significant number of digits of the quantization coefficient is output in step S83, in step S89, in the stored value of the variable x When it is 0, the maximum significant digit calculation unit 63 sets the value of the stored variable Binit to Binit=B.

That is, in the case where the value of the stored variable x is 0, the maximum significant number of digits calculation unit 63 indicates the maximum significant number of quantization coefficients on the row (y-1) stored therein first by indicating the maximum significant The value of the variable Binit of the number of bits is used as the value of the variable B indicating the maximum significant number of bits of w quantized coefficients encoded last time, thereby updating the variable Binit.

Thus, in the case of the variable x=0, the variable Binit value is set to Binit=B, so that the number of maximum significant digits of the w quantization coefficients starting from x=0 in the previous line (for example, line y) can be used In related relation, encoding of quantized coefficients is performed on w quantized coefficients starting from x=0 in the next row (for example, row (y+1)).

In step S90, the significant digit extracting unit 65 takes a predetermined variable i, and changes the variable i from 0 to (w-1) to extract the significant value of the quantized coefficient from the quantized coefficient at the position (x-i, y) on the row y. digits, where these quantization coefficients have been provided from the quantization unit 22. The significant digit extraction unit 65 supplies (the data of) the extracted significant digits of the quantization coefficients to the VLC encoding unit 66 and the sign extraction unit 67 . Also, the VLC encoding unit 66 outputs codes indicating the absolute values of w quantization coefficients to the code linking unit 69 based on the significand digits supplied from the significand digit extraction unit 65 (encodes the significand digits).

Now, the value of x at position (x+i, y) is the value of variable x stored at the most significant digit calculation unit 63 . For example, in the case where the value of the variable x stored at the maximum significant digit calculation unit 63 is 0, the value of the variable B stored at the significant digit extraction unit 65 is 3, and further, the quantization unit 22 extracts Unit 65 provides w (four) quantized coefficients "-0101", "+0011", "-0110", and "+0010" shown in FIG. 3, which correspond to each position (x+i, y )(0≤i≤3), that is, quantized coefficients at positions (0, y), (1, y), (2, y) and (3, y), the significant digit extracting unit 65 extracts significant digits from these quantized coefficients digit.

In this case, the value of the variable B stored at the significant-digit extraction unit 65 is 3, and the significant-digit number is 3 digits, so the significant-digit extraction unit 65 selects from the quantization coefficient “- 0101" is extracted as the value "101" of three digits starting from the lowest order.

In the same manner, the significant digit extracting unit 65 extracts from the quantized coefficients "+0011", "-0110" at the position (x+1, y), the position (x+2, y), and the position (x+3, y). , and "0010" sequentially extract the values "011", "110", and "010", which are three digits starting from their lowest order. Thus, the (codes) of the significant digits "101", "011", "110", and "010" of the quantization coefficients "-0101", "+0011", "-0110", and "+0010" are from the significant digit The extraction unit 65 is output to the VLC encoding unit 66 and the symbol extraction unit 67 . The VLC encoding unit 66 encodes the codes "101", "011", "110", and "010" supplied from the significant-digit extracting unit 65, and outputs the codes "101011110010" indicating the absolute values of w quantization coefficients to Code Link Unit 69.

In step S91, the sign extracting unit 67 takes a predetermined variable i, and changes the variable i from 0 to (w-1) so as to start from the position on the row y of which the absolute value of the quantization coefficient is not 0 supplied from the quantization unit 22 The quantization coefficient at (x+i, y) extracts the sign of the quantization coefficient, and supplies (the data of) the extracted sign to the VLC encoding unit 68 . The VLC encoding unit 68 encodes the sign from the sign extracting unit 67 , and outputs a code indicating the sign of the quantized coefficient thus obtained to the code linking unit 69 .

When the code indicating the sign of the quantized coefficient is input from the VLC encoding unit 68, the code linking unit 69 links the indications respectively supplied from the VLC encoding unit 62, the VLC encoding unit 64, the VLC encoding unit 66, and the VLC encoding unit 68 whether the A code in which all the quantization coefficients of a row are 0, a code indicating the maximum number of significant digits of the quantization coefficient, a code indicating the absolute value of the quantization coefficient, and a code indicating the sign of the quantization coefficient, the linked codes are output as an encoded image, The w-set encoding process ends, the flow returns to step S50 in FIG. 6, and the processes from step S51 are executed.

Now, the value of x at position (x+i, y) is the value of variable x stored at the most significant digit calculation unit 63 . For example, the value of the variable x stored at the maximum significant digit calculation unit 63 is 0, and the quantization unit 22 provides a value corresponding to each position (x+i, y) (0≤i≤3) shown in FIG. 3 That is, w (four) quantization coefficients "-0101", "+0011" corresponding to the quantization coefficients at positions (0, y), (1, y), (2, y), and (3, y) , "-0110", and "+0010", "-0101", "+0011", "-0110", and "+0010" are all non-zero, so the sign extraction unit 67 extracts from these quantization coefficients Extract symbols.

In this case, the sign extracting unit 67 extracts the sign "-" of the quantization coefficient corresponding to the position (x, y) from the quantization coefficient "-0101".

In the same manner, the sign extracting unit 67 sequentially selects from the quantization coefficients "+0011", "-0110" corresponding to the positions (x+1, y), (x+2, y), and (x+3, y). , "+0010" to extract the symbols "+", "-", and "+" of the quantization coefficient. Thus, the signs "-", "+", "-", "+" of the quantization coefficients "-0101", "+0011", "-0110", and "+0010" are output from the sign extraction unit 67 to the VLC encoding Unit 68. The VLC encoding unit 68 encodes the sign “−”, “+”, “−”, “+” of the quantization coefficient supplied from the sign extracting unit 67 .

For example, the VLC encoding unit 68 outputs code 1 when a sign "-" is input, and outputs code 0 when a sign "+" is input, thereby encoding the input sign. In this case, the signs "-", "+", "-", and "+" of the quantized coefficients are input to the VLC encoding unit 68, so the VLC encoding unit 68 will be represented by the codes "1", "0", " The code "1010" composed of 1", "0" is output to the code linking unit 69 as a code indicating the sign of the quantization coefficient.

Thus, the entropy encoding unit 23 encodes the quantization coefficients of the subbands in batches in predetermined units, and outputs a code indicating the maximum significant number of digits of the quantization coefficient, a code indicating the absolute value of the quantization coefficient, and a sign indicating the quantization coefficient code.

Thus, encoding the quantized coefficients of subbands in batches in predetermined units, unlike the case of encoding an image with, for example, JPEG 2000, eliminates the need to perform processing multiple times on the bit plane based on a plurality of encoding passes, and performs Variable code length encoding, so the amount of processing used for encoding can be significantly reduced. Therefore, image encoding can be performed at a higher speed, and an encoding device for encoding a high-resolution image in real time can be realized inexpensively.

Furthermore, there is no need to explicitly encode the length of the code at the image encoding device 11 in the case of encoding an image, so the amount of code can be reduced, and there is no need to manage information on the length of the code.

Note that although it has been described in the foregoing that, among the w quantization coefficients, the significant digit of the quantization coefficient having the largest absolute value is adopted as the value of the variable Bnew indicating the maximum significant digit, however, the value of the variable Bnew is equal to or A value larger than the significant number of digits of the quantization coefficient having the largest absolute value among the w quantization coefficients is sufficient. If the value of the variable Bnew is larger, the code amount increases for the code indicating the absolute value of the quantization coefficient, but the increase in the code amount of the code indicating the absolute value of the quantization coefficient can be achieved by setting the value of the variable Bnew equal to or a value larger than the number of significant bits of the quantization coefficient with the largest absolute value.

Next, description will be made regarding an image decoding device for decoding an image encoded by the image encoding device 11 .

Fig. 8 is a block diagram showing a configuration example of an image decoding device.

The image decoding device 111 is configured with an entropy decoding unit 121 , an inverse quantization unit 122 , and a wavelet inverse transformation unit 123 , and the encoded image (data) is input to the entropy decoding unit 121 .

The entropy decoding unit 121 performs entropy decoding on the code as an input encoded image, and supplies the quantized coefficient thus obtained to the inverse quantization unit 122 .

The inverse quantization unit 122 performs inverse quantization on the quantized coefficient supplied from the entropy decoding unit 121 , and supplies the wavelet coefficient for each subband obtained by the inverse quantization to the wavelet inverse transform unit 123 .

The wavelet inverse transform unit 123 subjects the wavelet coefficients for each subband supplied from the inverse quantization unit 122 to wavelet inverse transform, and outputs an image obtained as a result thereof as a decoded image.

Also, the entropy decoding unit 121 of the image decoding device 111 that performs such processing is configured in more detail, for example, as shown in FIG. 9 .

In more detail, entropy decoding unit 121 has code division unit 151 , row determination unit 152 , generation unit 153 , VLC decoding unit 154 , VLC decoding unit 155 , VLC decoding unit 156 , quantization coefficient synthesis unit 157 , and switching unit 158 .

The code dividing unit 151 divides a code which is an input encoded image based on information supplied from each of the line determining unit 152, the VLC decoding unit 154, the VLC decoding unit 155, and the VLC decoding unit 156, and A division code of a predetermined length is supplied to the row determination unit 152 , the VLC decoding unit 154 , the VLC decoding unit 155 , or the VLC decoding unit 156 .

That is, the code division unit 151 divides the input code into a code indicating whether the quantized coefficients of one encoded row are all 0, a code indicating the maximum significant number of digits of w encoded quantized coefficients, a code indicating w encoded quantized coefficients The code of the absolute value of , and the code indicating the sign of the encoded quantization coefficient are supplied to each of the row determination unit 152, the VLC decoding unit 154, the VLC decoding unit 155, and the VLC decoding unit 156, respectively.

The line determination unit 152 determines whether or not the quantization coefficients of one line of the encoded subband are all 0 based on the code supplied from the code division unit 151, and supplies information indicating the result of the determination to the code division unit 151, the generation unit 153, and the VLC decoding unit 154 .

The generating unit 153 generates a code indicating that the quantization coefficient equivalent to one row is 0 based on the information indicating the determination result from the row determining unit 152 , and supplies it to the switching unit 158 .

The VLC decoding unit 154 decodes the code indicating the maximum significant digits of the w coded quantization coefficients supplied from the code division unit 151, obtains the maximum significant digits of the w coded quantization coefficients, and will indicate the obtained maximum significant digits The information on the number of bits is supplied to the code division unit 151 , the VLC decoding unit 155 , and the quantization coefficient synthesis unit 157 .

The VLC decoding unit 155 decodes the code indicating the absolute value of the quantized coefficient supplied from the code dividing unit 151 based on the information indicating the maximum significant digit from the VLC decoding unit 154, and converts the significant digits of the w quantized coefficients thus obtained (the data of) is supplied to the VLC decoding unit 156 and the quantization coefficient synthesis unit 157 . And, the VLC decoding unit 155 supplies information representing the result of decoding the code indicating the absolute value of the quantized coefficient to the code dividing unit 151 .

Based on the significant digit of the quantized coefficient supplied from the VLC decoding unit 155, the VLC decoding unit 156 decodes the code indicating the sign of the quantized coefficient supplied from the code dividing unit 151, and supplies (data of) the sign of the quantized coefficient thus obtained to the quantization coefficient synthesis unit 157 . And, the VLC decoding unit 156 supplies information indicating the result of decoding the code indicating the sign of the quantized coefficient to the code dividing unit 151 .

The quantized coefficient synthesizing unit 157 synthesizes the significant digit of the quantized coefficient supplied from the VLC decoding unit 155 and the sign of the quantized coefficient supplied from the VLC decoding unit 156 based on the information indicating the maximum significant digit supplied from the VLC decoding unit 154, and will The obtained w quantization coefficients are supplied to the switching unit 158 .

The switching unit 158 outputs the quantization coefficient from the generation unit 153 or the quantization coefficient synthesis unit 157 .

FIG. 10 is a block diagram showing a more detailed configuration of the code division unit 151 .

The code division unit 151 has a control unit 171 and a memory 172 . When a code as an encoded image is input, the control unit 171 supplies the input code to the memory 172 for temporary storage.

Then, the control unit 171 selects from the information temporarily stored in the memory 172 based on the information supplied from each of the row determination unit 152, the VLC decoding unit 154, the VLC decoding unit 155, and the VLC decoding unit 156 shown in FIG. A code of a predetermined length is read out of the code, and supplied to the row determination unit 152 , the VLC decoding unit 154 , the VLC decoding unit 155 , or the VLC decoding unit 156 .

Also, the code division unit 151 may be configured as shown in FIG. 11 in addition to the configuration example shown in FIG. 10 .

The code division unit 151 shown in FIG. 11 has a control unit 191, a switch 192, and nodes 193-1 to 193-4.

When a code that is an encoded image is input from the code dividing unit 151, the control unit 191 bases on each of the line determining unit 152, the VLC decoding unit 154, the VLC decoding unit 155, and the VLC decoding unit 156 shown in FIG. Information supplied in one, the switch 192 is controlled, and a code of a predetermined length is supplied to the row determination unit 152 , the VLC decoding unit 154 , the VLC decoding unit 155 , or the VLC decoding unit 156 .

That is, the nodes 193-1 to 193-4 are each respectively connected to the row determination unit 152, the VLC decoding unit 154, the VLC decoding unit 155, and the VLC decoding unit 156, and the control unit 191 selects the nodes 193-1 to 193 One of -4 serves as the supply destination of the code, and controls the connection between the switch 192 and the selected node.

The switch 192 connects the node selected based on the control of the control unit 191 to the input, so the code input to the code division unit 151 is supplied to the row determination unit 152 and the VLC decoding unit via the switch 192 and the nodes connected to the switch 192 154. The VLC decoding unit 155, or the VLC decoding unit 156 serves as a code supply destination.

Next, decoding processing by the image decoding device 111 will be described with reference to the flowchart shown in FIG. 12 . This decoding process starts when a code as an encoded image is input to the entropy decoding unit 121 .

In step S131 , the entropy decoding unit 121 performs entropy decoding processing, performs entropy decoding of the code that has been input as an encoded image, and supplies the quantized coefficient thus obtained to the inverse quantization unit 122 . Although the details of the entropy decoding process will be described in detail below, by this entropy decoding process, entropy decoding unit 121 decodes quantized coefficients at consecutive positions on one line of encoded subbands w at a time, and decodes the decoded quantized coefficients The coefficients are supplied to the inverse quantization unit 122 .

In step S132 , the inverse quantization unit 122 performs inverse quantization on the quantized coefficients supplied from the entropy decoding unit 121 , and supplies the wavelet coefficients of each subband obtained by the inverse quantization to the wavelet inverse transform unit 123 .

In step S133 , the wavelet inverse transform unit 123 subjects the wavelet coefficients of each subband supplied from the inverse quantization unit 122 to wavelet inverse transform, and outputs an image obtained as a result thereof, whereby the decoding process ends.

Thus, the image decoding means 111 decodes and outputs the encoded image.

Next, entropy decoding processing corresponding to the processing of step S131 in FIG. 12 will be described with reference to the flowchart in FIG. 13 .

In step S161, the row determination unit 152 takes the variable y indicating the row of the subband to be decoded now as y=0, and stores it.

In step S161, the VLC decoding unit 154 fetches a variable Binit indicating the maximum number of significant digits of w quantization coefficients input first on one line (y−1) preceding the line y indicated by the variable y stored in the line determining unit 152. is Binit=0, and it is stored.

For example, in the case where row (y-1) is L1 shown in FIG. 2 , the value of the variable Binit indicating the maximum number of significant digits of w quantization coefficients input first on row (y-1) is obtained from The maximum significant number of bits of the w quantization coefficients at the left edge of line L1 in the figure. And, in the case where the variable y stored in the row determination unit 152 is y=0, the row (y−1) does not exist, so the value of the variable Binit is Binit=0.

And, in step S162 , the code division unit 151 supplies the first 1-bit code of the input code to the row determination unit 152 as a code indicating whether all quantization coefficients of the row to be decoded now are 0.

In step S163, line determination unit 152 determines whether the 1-bit code read in (supplied) from code division unit 151 is 0, generates information indicating the determination result, and supplies to generation unit 153, VLC decoding unit 154, and code Segmentation unit 151 .

In a case where it is determined in step S163 that the code is 0, this means that the quantization coefficients of row y are all 0, so the row determination unit 152 advances the flow to step S164. In step S164 , the generation unit 153 takes all quantization coefficients on the row y to be 0 based on the information indicating the determination result from the row determination unit 152 . Then, the generating unit 153 generates a code indicating the quantization coefficient of the row y, and supplies it to the switching unit 158 .

For example, in the case where one quantization coefficient is represented by four digits as shown in FIG. 3 and there are five quantization coefficients in one row, the generation unit 153 generates 20 (=4×5) zeros as the quantization coefficient indicating row y and provide it to the switching unit 158. The switching unit 158 outputs 20 consecutive 0s supplied from the generating unit 153 to the inverse quantization unit 122 as a code indicating the quantization coefficient of one line.

In step S165 , the VLC decoding unit 154 sets the value of the variable Binit stored therein to Binit=0 based on the information indicating the determination result from the line determination unit 152 , and updates the variable Binit.

In step S166, the line determination unit 152 determines whether there is an unprocessed line in the decoded subband. That is, the line determination unit 152 determines whether quantized coefficients at positions on all lines of the decoded subband have been decoded.

In a case where it is determined in step S166 that there is an unprocessed row, the row determination unit 152 advances the flow to step S167 to perform a processing on the next row (y+1) of the row y indicated by the variable y stored in itself. The quantized coefficients at each position of the are encoded.

In step S167, the row determination unit 152 increments the variable y indicating the stored row by y=y+1, returns the flow to step S163, and performs subsequent processing.

On the contrary, in the case where it is determined in step S166 that there are no unprocessed lines, the quantized coefficients of all the lines constituting the subband have been decoded, and therefore the line determination unit 152 ends this entropy decoding process, and the flow returns to FIG. 12 . Step S131, and the processing from step S132 onwards is performed.

And, in a case where it is determined in step S163 of FIG. 13 that the code is not 0, the row determination unit 152 advances the flow to step S168. In step S168, based on the information indicating the determination result from the row determination unit 152, the VLC decoding unit 154 sets the value of the variable x indicating the value to be decoded now to x=0 and stores the variable x. The x-coordinate of the position on row y of the first quantization coefficient to be input among the w quantization coefficients.

And, in step S168, the VLC decoding unit 154 takes the value of the variable B as B=Binit, and stores the variable B indicating the maximum significant number of bits of the w quantization coefficients decoded last time. That is, the VLC decoding unit 154 updates the value of the variable B, makes the value of the variable B the stored value of the variable Binit, and stores the updated value of the variable B.

Further, in step S168, the code division unit 151 supplies the next 1-bit code of the input code to the VLC decoding unit 154 as an indication of the w numbers to be decoded from now on, based on the information indicating the determination result from the row determination unit 152. The code of whether the maximum significant digit of the quantization coefficient is changed.

In step S169, the entropy decoding unit 121 performs set decoding processing. Although details of the w-set decoding process will be described later, in the w-set decoding process, the entropy decoding unit 121 decodes w consecutive quantized coefficients on the row y indicated by the variable y stored in the row determining unit 152 .

In step S170, the VLC decoding unit 154 determines whether there is an unprocessed quantization coefficient on the row y. That is, the VLC decoding unit 154 determines whether or not all quantized coefficients stored at positions on the line y indicated by the variable y at the line determining unit 152 have been decoded.

In a case where it is determined in step S170 that there are unprocessed quantized coefficients on the line y, the VLC decoding unit 154 advances the flow to step S171 to decode the next w quantized coefficients.

In step S171, the VLC decoding unit 154 takes the variable x stored therein as x=x+w, and returns the flow to step S169. Therefore, the quantization coefficient at each of the positions (x+w, y), (x+w+1, y), ..., (x+2w-1, y) on the row y is determined in the following step S169 is encoded in the processing.

And, in the case where it is determined in step S170 that there is no unprocessed quantized coefficient on line y, the quantized coefficients at all positions on line y have already been decoded, so the VLC decoding unit 154 returns the flow to step S166, and Perform follow-up processing.

Thus, the entropy decoding unit 121 decodes the quantized coefficients at each position of the subband in raster scan order every predetermined number.

Thus, by decoding quantized coefficients at each position of a subband in raster scan order every predetermined number, input quantized coefficients can be processed in the order in which they are input, and delay due to quantized coefficient decoding can be reduced .

Next, w-set decoding processing corresponding to the processing of step S169 in FIG. 13 will be described with reference to the flowchart shown in FIG. 14 .

As described above, in step S168 of FIG. 13 , the code division unit 151 to the VLC decoding unit 154 are supplied with a 1-bit code indicating whether the maximum significant digits of the w quantization coefficients to be decoded are changed.

In step S211 of FIG. 14 , the VLC decoding unit 154 determines whether the read-in (supplied) 1-bit code is 0.

In the case where it is determined in step S211 that the read-in code is 0, the maximum significant digit has not changed, so the VLC decoding unit 154 generates information meaning that the maximum significant digit has not changed, and supplies it to the code dividing unit 151, VLC The decoding unit 155, and the quantization coefficient synthesis unit 157, skip the processing in steps S212 to S214, and advance the flow to step S215.

That is, in the case where the code indicating whether the maximum significant digit is changed is 0, as described with reference to FIG. Is the code to increase or decrease and the code indicating the amount of change in the maximum significant digit, but the code indicating the absolute value of the quantization coefficient, so skip as a code for decoding indicating whether the maximum significant digit is increased or decreased Step S212 to Step S214 of the code and the code indicating the change amount of the maximum significant digit.

On the contrary, in the case where it is determined that the read-in code is not 0 in step S211, the maximum significant number of digits has changed, so the VLC decoding unit 154 advances the flow to step S212, reads in a 1-bit code from the code dividing unit 151, and stores Its value serves as a predetermined variable m.

In step S213, VLC decoding unit 154 reads codes from code dividing unit 151 until code is 1 (until code 1 is read), and stores the number of code 0s read up to that point as variable n. For example, in the case where the third code read in by the VLC decoding unit 154 from the code dividing unit 151 is 1, that is, in the case where the code “001” is read in by the VLC decoding unit 154, until the VLC decoding unit 154 The number of code 0s read until 1 is read is 2, so the VLC decoding unit 154 stores 2, which is the number of code 0s read, as the value of the variable n.

In step S214 , the VLC decoding unit 154 obtains the value of B indicating the maximum significant number of digits from the following expression (2), and stores the obtained value of the variable B.

B＝B+(n+1)×(-1)^m ....(2)

Here, the left side of the expression (2) represents the value of the variable B to be newly obtained, and the B on the right side represents the stored value of the variable B. Also, the symbol "^" in Expression (2) represents an exponent. Thus, (-1)^m means (-1) raised to the mth power.

The VLC decoding unit 154 calculates Expression (2) based on the stored variable B, variable m, and variable n, and changes the stored variable B. When updating variable B indicating the maximum number of significant digits, VLC decoding unit 154 generates information indicating the updated maximum significant number of digits, and supplies it to code division unit 151 , VLC decoding unit 155 , and quantization coefficient synthesis unit 157 .

When a new maximum significant digit is obtained in step S214 or when it is determined in step S211 that the 1-bit code that has been read in is 0, the VLC decoding unit 154 advances the process to step S215, and when the value of the stored variable x is In the case of 0, the value of the stored variable Binit is set to Binit=B.

That is, in the case where the stored value of the variable x is 0, the VLC decoding unit 154 stores the variable Binit indicating the maximum number of significant digits of the w quantization coefficients input first on the line (y-1). The value of is taken as the value of the variable indicating the maximum number of significant bits of w quantized coefficients to be decoded from now on, and the variable Binit is updated.

Thus, in the case of the variable x=0, the value of the variable Binit is set to Binit=B, so that the maximum significant number of bits corresponding to w quantization coefficients starting from x=0 in the previous line (for example, line y) can be used , the decoding of the quantized coefficients is performed on w quantized coefficients starting from x=0 in the next row (for example, row (y+1)).

In step S216, the VLC decoding unit 155 takes a predetermined variable i, and changes the variable i from 0 to (w-1) so as to read the code from the code division unit 151 in increments of B bits, and converts the read-in code The b bits of are supplied (output) to the VLC decoding unit 156 and the quantization coefficient synthesis unit 157 as a code indicating the significant digit of the quantization coefficient at the position (x+i, y) on the row y. And, the VLC decoding unit 155 generates information indicating the significant digit of the quantization coefficient, and supplies it to the code dividing unit 151 .

Here, the value of x at the position (x+i, y) is the value of the variable x stored at the VLC decoding unit 154 . For example, in the case where the value of the variable x stored at the VLC decoding unit 154 is 0, and the value of the variable B stored at the VLC decoding unit 155 is 3, the VLC decoding unit 155 changes from The code division unit 151 reads in a 3-bit code, and outputs the read-in 3-bit code as the significant digit of the quantization coefficient at position (0, y).

In the same manner, the VLC decoding unit 155 reads in another 3-bit code from the code division unit 151 in the state of variable i=1, and outputs the read-in 3-bit code as quantization at position (1, y) Significant digits of the coefficient, read in the next 3-bit code from the code division unit 151 in the state of the variable i=2, and output the read-in 3-bit code as the quantized coefficient at the position (2, y) significant digit, and read in the next 3-bit code from the code division unit 151 under the state of the variable i=3, and output the read-in 3-bit code as the effective value of the quantization coefficient at the position (3, y) digit.

In step S217, the VLC decoding unit 156 takes a predetermined variable i, and changes the variable i from 0 to (w-1), and the significant digit of the quantization coefficient at the position (x+i, y) on the row y ( Absolute value) is not 0, a 1-bit code is read from code division section 151 . The VLC decoding unit 156 then decodes the code that has been read in, and supplies (outputs) the code thus obtained to the quantization coefficient synthesis unit 157 as a sign of the quantization coefficient. And, the VLC decoding unit 156 generates information indicating the sign of the quantized coefficient, and supplies it to the code dividing unit 151 .

Here, the value of x at the position (x+i, y) is taken as the value of the variable x stored at the VLC decoding unit 154 . For example, in the case where the value of the variable x stored at the VLC decoding unit 154 is 0, and (indicating) non-zero significant digit (code) is supplied from the VLC decoding unit 155, the VLC decoding unit 156 takes the variable i= 0, and a 1-bit code is read in from the code division unit 151, and when the code is 0, the code indicating the sign “-” of the quantization coefficient at the position (0, y) is supplied to the quantization coefficient synthesis unit 157, And in the case where the code is 1, the code indicating the sign “+” of the quantization coefficient at the position (0, y) is supplied to the quantization coefficient synthesis unit 157 .

Also, in the case where the absolute value of (the code of) the significant digit supplied from the VLC decoding unit 155 is 0, there is no sign for the quantized coefficient at the position (0, y), so the VLC decoding unit 156 does not read from The code dividing unit 151 reads the code.

In the same manner, in the case where (the absolute value of) the significant digit supplied from the VLC decoding unit 155 is not 0, the VLC decoding unit 156 takes the variable i=1, and reads in a 1-bit code from the code dividing unit 151, and in When this code is 0, a code indicating a sign "-" is supplied to the quantization coefficient synthesis unit 157, and when this code is 1, a code indicating a sign "+" is supplied.

Furthermore, in the case where the significant digit next supplied from the VLC decoding unit 155 is not 0, the VLC decoding unit 156 takes the variable i=2, and reads in a 1-bit code from the code division unit 151, and when the code is 0 Here, a code indicating a sign "-" is supplied to the quantized coefficient synthesis unit 157, and in a case where the code is 1, a code indicating a sign "+" is supplied. And, in the case that the significant digit provided next from the VLC decoding unit 155 is not 0, the VLC decoding unit 156 takes the variable i=3, and reads in a 1-bit code from the code division unit 151, and when the code is 0 Here, a code indicating a sign "-" is supplied to the quantized coefficient synthesis unit 157, and in a case where the code is 1, a code indicating a sign "+" is supplied.

In step S218, the quantized coefficient synthesis unit 157 synthesizes the significant digit supplied from the VLC decoding unit 156 and the sign supplied from the VLC decoding unit 155, outputs the quantized coefficient thus obtained to the inverse quantization unit 122 via the switching unit 158, and ends w set The decoding process returns the flow to step S169 in FIG. 13 , and the processes from step S170 are executed.

For example, the number of bits of the absolute value of the quantization coefficient to be output is determined in advance. In the case where the number of digits of the absolute value of the quantized coefficient determined in advance to be output is 4 digits, and the maximum significant digit indicated by the information indicating the maximum significant digit from the VLC decoding unit 154 is 3, the valid The digit "101" is supplied from the VLC decoding unit 155, and when a code indicating a sign "-" is supplied from the VLC decoding unit 155, the quantization coefficient synthesis unit 157 outputs the quantization coefficient "-0101".

That is, the number of digits of the absolute value of the quantization coefficient to be output is 4 digits, and the significant digit is "101" (three digits), so for the absolute value of 4 digits of the quantization coefficient of "0101", The quantization coefficient synthesis unit 157 takes one higher-order bit of the significant digit "101" as 0. Furthermore, the sign "-" of the quantization coefficient and the absolute value "0101" of the quantization coefficient are synthesized to obtain "-0101", which is output as the quantization coefficient.

Note that in the case where the significant digit supplied from the VLC decoding unit 155 is 0, the quantized coefficient synthesis unit 157 outputs an unsigned quantized coefficient. For example, the number of digits of the absolute value of the quantization coefficient to be output determined in advance is 4 digits, and the maximum significant digit indicated by the information indicating the maximum significant digit from the VLC decoding unit 154 is 3, and from In the case where the VLC decoding section 155 provides the significant digit "000", the quantization coefficient synthesis section 157 outputs the quantization coefficient "0000".

Thus, the entropy decoding unit 121 decodes quantized coefficients of subbands in batches of predetermined units.

Thus, unlike the case of decoding an image with JPEG, decoding the encoded quantized coefficients of the subbands in batches in predetermined units eliminates the need to perform processing multiple times on the bit plane based on a plurality of encoding paths, and thus can be more Perform image decoding quickly. Therefore, a decoding device for decoding high-resolution images in real time can be realized inexpensively.

Here, for the above-mentioned image encoding device 11, it has been described that when encoding (or decoding) the absolute value of the quantized coefficient, the absolute value of the predetermined w quantized coefficients is sequentially encoded, but using the general-purpose DSP (Digital Signal processor) or a SIMD (Single Instruction Multiple Data) calculation command used together with a general-purpose CPU, by simultaneously (parallel) encoding (or decoding) w quantization coefficients, image encoding (or decoding) can be performed faster.

Here, examples of SIMD operation instructions include, for example, MMX (Multimedia Extensions), SSE (Streaming SIDM Extensions), SSE2, SSE3, etc. used with CPUs of Intel Corporation.

In the case of encoding the absolute value of the quantization coefficient using such a SIMD operation instruction, the entropy encoding unit 23 of the image encoding device 11 is configured as shown in FIG. 15 , for example.

The entropy coding unit 23 shown in FIG. 15 is the same as the entropy coding unit 23 shown in FIG. 64. Significant digit extracting unit 65, VLC encoding unit 66, sign extracting unit 67, and VLC encoding unit 68, and the difference is that a buffer 301 is provided for code linking unit 69. Note that parts shown in FIG. 15 corresponding to parts in FIG. 4 are denoted by the same reference numerals, and descriptions thereof will be omitted.

The buffer 301 of the code linking unit 69 temporarily stores codes indicating whether or not all quantization coefficients of one line are 0, indicating the maximum A code indicating the number of significant digits, a code indicating the absolute value of the quantization coefficient, and a code indicating the sign of the quantization coefficient.

The storage area of the buffer 301 is managed in increments of 32 bits, and codes (data) input to the buffer 301 are stored divided into codes for scalar calculation processing and codes for vector calculation processing. That is, a 32-bit storage area stores codes for scalar calculation processing or codes for vector calculation processing as codes (data) to be temporarily stored.

With the entropy encoding unit 23 shown in FIG. 15 , the absolute value of the quantization coefficient is encoded in parallel using SIMD operation instructions, so the code indicating the absolute value of the quantization coefficient is taken as the code for vector calculation processing, and the other codes Taken as the code for scalar computation processing.

Note that in the following description, among the 32-bit storage areas provided to the buffer 301, a storage area in which codes for scalar calculation processing are stored will be referred to as a scalar area, and a code for vector calculation processing is stored therein. The storage area of the code will be called the vector area.

Next, entropy encoding performed by the entropy encoding unit 23 shown in FIG. 15 will be described with reference to FIG. 16 .

For example, it can be said that the 12 quantization coefficients "-0101", "+0011", "-0110", "+0010", "+0011", " +0110”, “0000”, “−0011”, “+1101”, “−0100”, “+0111”, and “−1010” are input to the entropy encoding unit 23 .

In the same manner as described with reference to FIG. 3 , the code linking unit 69 of the entropy encoding unit 23 is supplied with a code "1" indicating whether or not the quantization coefficients of the row to be encoded are all 0, and a code "10001" indicating that the row to be encoded is all 0. The maximum significant digits of the four input quantization coefficients "-0101", "+0011", "-0110", and "+0010" of one.

Next, a code "110001" consisting of a code "1" indicating whether or not the quantization coefficients of the row to be encoded are all 0 and a code "10001" indicating the maximum significant number of digits of the quantization coefficient is stored in the buffer of the code linking unit 69. In the 32-bit scalar area set in the register 301, as indicated by the arrow A11.

In the example shown in FIG. 16 , the codes to be stored in the scalar area are stored in order from higher-order bits in the left-to-right direction in the drawing. When codes are stored in the entirety of one scalar area, that is, when 32-bit codes are stored in one scalar area, a new scalar area is set in the buffer 301, and codes used together with scalar calculation processing are stored sequentially in the newly set scalar area.

When code "110001" consisting of code "1" indicating whether or not the quantization coefficients of the row to be encoded are all 0 and code "10001" indicating the maximum significant number of digits of the quantization coefficient is stored in the scalar area, next, The entropy encoding unit 23 simultaneously stores (arranged in parallel) in the vector area a number representing each of the w (four) quantization coefficients "-0101", "+0011", "-0110", and "+0010" input first. A code equivalent to the most significant digit of the absolute value.

The maximum significant number of digits of the quantization coefficients "-0101", "+0011", "-0110", and "+0010" is "3", as described with reference to 3, thus representing the codes of the absolute values of the four quantization coefficients They are "101", "011", "110", and "010" respectively, and as shown by arrow A12, the codes "101", "011", "110", and "010" representing the absolute value of the quantization coefficient The arrangement is stored in a single vector area set to the buffer 301 .

Now, the vector area is further divided into four 8-bit areas, and in each of the four areas of the vector area in the left-to-right direction in the figure, stored in the order starting from the higher-order bits are the bits representing the same length (bit length) codes for the absolute values of the four quantized coefficients.

For the vector area shown by the arrow A12, in the figure, code "101" indicating the absolute value of the quantization coefficient is stored in the left 8-bit area from the left, and code "011" is stored in the left 8-bit area from the left. In the second 8-bit area from the left, the code "110" is stored in the second area from the right from the left, and the code "010" is stored in the rightmost area from the left.

And, in the same manner as the case of the scalar area, for the vector area, when the code is stored in the whole of one vector area, that is, when the 32-bit code is stored in one vector area, a new vector area is set In the buffer 301, codes used together with vector calculation processing are sequentially stored in the newly set vector area.

When codes indicating the absolute values of quantization coefficients "-0101", "+0011", "-0110", and "+0010" are stored in the vector area, the entropy encoding unit 23 stores in the scalar area representing four quantization The code for the sign of the coefficient, as indicated by arrow A13.

As indicated by arrow A11, code "110001" consisting of code "1" indicating whether or not the quantization coefficients of the row to be encoded are all 0 and code "10001" indicating the maximum number of significant digits of the quantization coefficient has been stored in the scalar area, therefore, as indicated by the arrow A13, the code "1010" indicating the sign of the quantization coefficients "-0101", "+0011", "-0110", and "+0010" is stored in the scalar area already stored To the right of the code "110001" (continuously to the right of the code "110001").

Furthermore, when the first four quantization coefficients are encoded, the entropy encoding unit 23 encodes the next four quantization coefficients "+0011", "+0110", "0000", and "-0011".

First, the entropy encoding unit 23 compares the maximum significant digits "3" of the four quantized coefficients encoded last time with the next four quantized coefficients "+0011", "+0110", and "0000" encoded this time. , "3" of the maximum significant digit of "-0011", and since the maximum significant digit has not changed, a code indicating that the maximum significant digit has not changed is stored in the scalar area contiguous to the right of the stored code "1100011010" "0", as a code for indicating the maximum significant number of digits, as indicated by arrow A14.

Next, as shown by the arrow A15, the entropy encoding unit 23 simultaneously stores each of the codes "011", "110", "000" and "011" in the vector area, which represent the w (four) input codes of this time. ) the absolute value of each of the quantization coefficients "+0011", "+0110", "0000", and "-0011".

As indicated by arrow A12, in the 8-bit area of the vector area on the left in the figure, in the second 8-bit area from the left, in the second 8-bit area from the right, and in the rightmost In the 8-bit area, the codes "101", "011", "110", and "010" have been stored, so as shown by the arrow A15, the entropy coding unit 23 can be used for each code "101" that has been stored in the vector area. , "011", "110", and each of the codes "011", "110", "000", and "011" are continuously stored on the right side of "010", which represent each quantization coefficient input this time the absolute value of .

Furthermore, the entropy encoding unit 23 takes the code "001" indicating the sign of the quantization coefficient whose absolute value is not 0 for the four quantization coefficients "+0011", "+0110", "0000", and "-0011" input this time, And it is stored consecutively to the right of the code "11000110100" already stored in the scalar area.

When the four quantization coefficients "+0011", "+0110", "0000", and "-0011" are encoded, the entropy coding unit 23 encodes the next four quantization coefficients "+1101", "-0100", "+ 0111", and "-1010" to perform encoding.

First, the entropy encoding unit 23 compares the maximum significant digits "4" of the four input quantization coefficients "+1101", "-0100", "+0111", and "-1010" with the four quantization coefficients encoded last time. The maximum significant digits "3" of the two quantized coefficients are compared, and as shown by arrow A17, a code "101" indicating the maximum significant digit is stored in the scalar area, which is replaced by "1" indicating that the maximum significant digit has changed. , "0" indicating the increase of the maximum significant digit, and "1" indicating the increase amount of the maximum significant digit.

In this case, the code "11000110100001" has been stored in the scalar area as indicated by arrow A16, so the entropy encoding unit 23 stores the code "101" indicating the maximum significant number of digits to the right of the code "11000110100001" in the drawing.

Furthermore, when codes indicating the maximum significant digits of the four quantization coefficients "+1101", "-0100", "+0111", and "-1010" are stored, as indicated by arrow A18, the entropy encoding unit 23 Each of the codes "1101", "0100", "0111", and "1010", which represent the absolute value of each of these quantization coefficients, is simultaneously stored in the vector area.

As indicated by arrow A15, in the 8-bit area of the vector area on the left in the figure, in the second 8-bit area from the left, in the second 8-bit area from the right, and in the 8-bit area on the far right bit areas, have stored the codes "101011", "011110", "110000", and "010011", respectively, so the 8-bit area on the left, from the second 8-bit area on the left, from the second 8-bit area on the right The first 8-bit area, and the rightmost 8-bit area, can only store two-bit codes.

Thereby, the entropy encoding unit 23 secures (provides) the new vector area in the buffer 301, as shown by the arrow A18, in the codes "101011", "011110", "110000", and "010011" already stored in the vector Codes "1101", "0100", "0111" representing the absolute value of the quantization coefficient input this time, and codes "11", "01" of the two higher-order bits of "1010" are successively stored on the right side of the ", "01", and "10", and in the 8-bit area of the newly set vector area in the figure (of the two vector areas indicated by arrow A18, the vector area at the bottom of the figure), from the 8th bit area on the left Two 8-bit areas, from the second 8-bit area on the right side, and the left side of each area in the rightmost 8-bit area, respectively store the code "1101" representing the absolute value of the quantization coefficient input this time ", "0100", "0111", and the two lower order bits "01", "00", "11", and "10" of "1010".

When codes indicating the absolute values of the four quantization coefficients "+1101", "-0100", "+0111", and "-1010" are stored, as indicated by arrow A19, the entropy encoding unit 23 is already in the scalar region On the right side of the code "11000110100001101" stored in , a code "0101" indicating the signs of four quantization coefficients whose absolute value is not zero is successively stored.

Thus, when the input quantization coefficient is encoded, the entropy encoding unit 23 sequentially outputs the code stored in the scalar area indicated by arrow A19, the upper vector area in the figure stored in the two vector areas indicated by arrow A19 The code in , and the code stored in the vector region below, as an encoded image.

In this case, no code is stored in the 11 bits on the right side of the figure in the scalar area shown in arrow A19. And, in the vector area at the bottom of the two vector areas indicated in arrow A19, the 8-bit area on the left, the second 8-bit area from the left, the second 8-bit area from the right, and the rightmost No code is stored in the right six-bit area in each of the 8-bit areas.

In the case of outputting the codes stored in the scalar area and the vector area in this way as an encoded image, there may be an area in which no code is stored in the scalar area and the vector area at the moment when the encoding of the input quantization coefficient is completed Next, for example, an arbitrary code such as a code "0" is stored, after which the codes stored in the scalar area and the vector area are encoded and output as an image.

Therefore, for example, in the example shown by arrow A19, the code "11000110100001101010100000000000" stored in the scalar area, the code "10101111011110011100000101001110" stored in the upper vector area in the figure, and the code stored in the lower vector area "01000000000000001100000010000000" are sequentially output as encoded images. Now, since the arbitrary codes stored in the scalar area and the vector area where no codes are stored at the time when quantization coefficient encoding is completed are not read during decoding, any kind of codes can be stored.

In the case of encoding the absolute value of the quantization coefficient also using the SIMD operation instruction, when an image is input, the image encoding device 11 executes the encoding process described with reference to the flowchart in FIG. 5 . And, at the entropy coding processing in FIG. 6 corresponding to step S13 in FIG. 6), the same processing as in the case of not using the SIMD operation instruction is performed, and for the w-set encoding processing corresponding to step S50, the processing different from that in the case of not using the SIMD operation instruction is performed.

The w-set encoding process in the case where the image encoding device 11 encodes the absolute value of the quantization coefficient using the SIMD operation instruction will be described below with reference to the flowchart in FIG. 17 . Note that each processing in step S311 to step S319 corresponds to each processing in step S81 to step S89 in FIG. 7 and is performed in the same manner. A description thereof is therefore redundant and thus omitted.

Also, in the case of encoding the absolute value of the quantization coefficient using a SIMD operation instruction, in the processing described with reference to FIGS. The code indicating whether the absolute values of the quantized coefficients are all 0, the code indicating the maximum number of significant digits of the quantized coefficient supplied from the VLC encoding unit 64 to the code linking unit 69, and the indication supplied from the VLC encoding unit 68 to the code linking unit 69 The codes of the signs of the quantization coefficients are each stored in the scalar area provided in the buffer 301 of the code linking unit 69, as described with reference to FIG. 16 .

In step S320, the significant digit extracting unit 65 selects the position (x, y), (x+1, y), . . . The significant digits of the quantized coefficients are simultaneously extracted from the w consecutive quantized coefficients. The significant digit extraction unit 65 supplies the extracted significant digits of the quantized coefficients to the VLC encoding unit 66 and the sign extraction unit 67 . And, the VLC encoding unit 66 simultaneously outputs codes indicating the absolute values of w quantization coefficients to the code linking unit 69 based on the significant digits supplied from the significant digit extraction unit 65 (encodes the significant digits).

Now, the value of x at the position (x, y) is taken as the value of the variable x stored at the maximum significant digit calculation unit 63, and the value of y is taken as the value of the variable y stored at the row determination unit 61. value. For example, in the case where the significant digit extraction unit 65 extracts the significant digits "101", "011", "110", and "010" as the significant digits of the quantization coefficients, codes "101" indicating the absolute values of the four quantization coefficients ", "011", "110", and "010" are supplied from the VLC encoding unit 66 to the code linking unit 69, so the code linking unit 69 encodes the code indicating the absolute value of the quantization coefficient supplied thereto, and as As indicated by arrow A12 in FIG. 16, it is stored in the vector area.

In step S321, the sign extracting unit 67 takes a predetermined variable i, and changes the variable i from 0 to (w-1) so that the quantization at the position (x+i, y) on the row y whose quantization coefficient is not 0 The signs of the quantized coefficients that have been supplied from the quantization unit 22 are extracted from the coefficients, and (the data of) the extracted signs are supplied to the VLC encoding unit 68 . Now, the value of x at the position (x, y) is taken as the value of the variable x stored at the maximum significant digit calculation unit 63, and the value of y is taken as the value of the variable y stored at the row determination unit 61. value.

The VLC encoding unit 68 encodes the sign from the sign extracting unit 67 , and outputs a code indicating the sign of the quantized coefficient thus obtained to the code linking unit 69 . Also, as described with reference to FIG. 16 , the code linking unit 69 stores the code indicating the sign of the quantization coefficient supplied from the VLC encoding unit 68 in the scalar area of the buffer 301 .

When the code indicating the sign of the quantization coefficient is stored in the scalar area of the buffer 301, the code linking unit 69 links the codes stored in the scalar area and the vector area of the buffer 301 as described with reference to FIG. 16 , and outputs the linked code as an encoded image, so that the w-set encoding process ends, and the process returns to step S50 in FIG. 6 , and the processes from step S51 are performed.

Thus, the entropy encoding unit 23 simultaneously encodes a predetermined number of absolute values of the quantization coefficients.

Entropy encoding by the JPEG 2000 method, based on multiple encoding paths, arithmetically encodes the quantized coefficients in increments of bit planes, so it is difficult to simultaneously perform predetermined processing in entropy encoding in parallel, but by the entropy encoding unit 23, it is not necessary to Bit-planes are added to perform complex processing, so the absolute values of multiple quantization coefficients can be encoded simultaneously.

Thus, simultaneously encoding the absolute values of a predetermined number of quantization coefficients enables simultaneous (parallel) execution of a plurality of processes, so that an image can be encoded at a higher speed.

Note that, in the process of step S321, it has been described that the encoding of the signs of w quantization coefficients is sequentially performed, but in the same manner as in the case of encoding the absolute value of the quantization coefficients, by using the SIMD operation instruction Encoding is performed simultaneously on the signs of w quantized coefficients. In this case, each of the codes indicating the signs of w quantization coefficients obtained by encoding is stored in the vector area of the buffer 301 which has been divided into w.

Also, although it has been described regarding the buffer 301 that one scalar area or vector area is a 32-bit area, and this 32-bit area is further divided into 8-bit areas for use, the size of the scalar area or vector area, etc. may be any size. For example, an arrangement may be made in which a scalar area or a vector area is set as a 128-bit area divided into eight 16-bit areas for use.

Furthermore, in the case of decoding an image encoded using a SIMD operation instruction, the code division unit 151 ( FIG. 9 ) for decoding the image of the image decoding device 111 is configured as shown in FIG. 10 , for example, and as an already Codes of encoded images are stored into the memory 172 32 bits at a time, as described with reference to FIG. 16 .

In the case of reading out the code from the memory 172 and outputting it, the control unit 171 first takes the storage area in which the first 32-bit code is stored as a scalar area, and reads and outputs an indication in order from the top of the scalar area. Whether the absolute values of the quantized coefficients of the row to be decoded are all 0, a code indicating the maximum number of significant digits of the quantized coefficient, or a code indicating the sign of the quantized coefficient.

And, in the case of reading out the code indicating the absolute value of the quantization coefficient from the memory 172, the control unit 171 fetches the 32-bit storage area following the storage area in the memory 172 as the scalar area (so that the code has not been read from this area) as a vector area, and a code indicating the absolute value of the quantization coefficient is read out from this vector area and output.

Note that, in the case where an image is encoded, the image is encoded so that at the moment when the code indicating the absolute value of the quantization coefficient is read out for the first time at the time of decoding, there is always a code indicating the absolute value of the quantization coefficient (code for vector calculation ) is stored in the next 32-bit storage area after the storage area used as the scalar area.

Furthermore, how many bits each storage area is divided into and stored in the memory 172 as the code of the coded image is based on a scalar area and a vector area in the case of encoding an image by the image encoding device 11. It is changed by how many bits are set. That is, the size of each of a plurality of storage areas into which the code as an encoded image is divided and stored in the memory 172 is the size of one scalar area and vector area in the case of encoding an image .

In the case of decoding the absolute value of the quantization coefficient also using the SIMD operation instruction, the image decoding device 111 executes the decoding process described with reference to the flowchart shown in FIG. 12 when an encoded image is input. And, for the entropy decoding process in FIG. 13 corresponding to the process of step S131 in FIG. 12 , for each of the processes in steps S161 to S168 in FIG. processing, the image decoding device 111 executes the same processing as in the case of not using the SIMD control command (the processing described with reference to FIG. 13 ), and in the w-set decoding processing corresponding to step S169, performs treated differently.

The w-set decoding process in the case where the image decoding device 111 decodes the absolute value of the quantization coefficient using the SIMD operation instruction is described below with reference to the flowchart of FIG. 18 .

Note that the processing in step S351 to step S355 corresponds to the processing in step S211 to step S215 in FIG. 14 , and each is performed in the same manner. Therefore, redundant description thereof will be omitted.

And, in the case of decoding the absolute value of the quantized coefficient using a SIMD operation instruction, a code as an image is stored, for example, in the memory 172 of the code division unit 151 divided into three 32-bit regions, as in FIG. 16 Shown by arrow A19. The row determination unit 152, the VLC decoding unit 154, and the VLC decoding unit 156 each take the uppermost area in FIG. 16 of the three 32-bit areas as a scalar area, and sequentially read out from the top of the scalar area (left side in the figure). And a code indicating whether or not the quantized coefficients of the row are all 0, a code indicating the maximum number of significant digits of the quantized coefficients, and a code indicating the sign of the quantized coefficients is decoded.

In step S356, the VLC decoding unit 155 simultaneously reads in the codes of B consecutive bits of w sets from the code division unit 151, and takes each of the B-bit codes of the read-in w sets as The codes of the significant digits of the quantization coefficients of w consecutive quantization coefficients at positions (x, y), (x+1, y), ..., (x+w-1, y) of , and provide (Output) to VLC decoding section 156 and quantization coefficient synthesis section 157 . And, the VLC decoding unit 155 generates information indicating the significant digit of the quantization coefficient, and supplies it to the code dividing unit 151 . Now, the value of x at the position (x, y) is taken as the value of the variable x stored at the VLC decoding unit 154 , and the value of y is taken as the value of the variable y stored at the row determination unit 152 .

For example, if we say that the predetermined number w is 4, the value of the variable B is 3, and the code as an image is stored in the memory 172 of the code division unit 151 divided into three 32-bit storage areas, as shown in FIG. 16 As shown by the arrow A19, the uppermost 32-bit storage area in Figure 16 has been used as a scalar area, wherein the code indicating whether all the quantization coefficients of the row are 0 and the code indicating the maximum significant number of quantization coefficients have been read out. code, and the code has not been read out from the next 32-bit storage area (the second storage area from the top), so the VLC decoding unit 155 takes the second storage area from the top as the vector area, and from the In the left 8-bit area of the vector area, the second 8-bit area from the left, the second 8-bit area from the right, and the rightmost 8-bit area, simultaneously read the indicated position (x, y) , (x+1, y), (x+2, y), and (x+3, y) the codes "101", "011", "110", and "010" of the significant digits of the quantization coefficients .

When codes indicating significant digits of w quantization coefficients are sequentially supplied to the VLC decoding unit 156 and the quantization coefficient synthesis unit 157, the processing of step S357 and the processing of step S358 are performed, but these processings are the same as the processing of step S217 in FIG. It is the same as the processing of step S218, so the description thereof will be omitted.

Thus, the entropy decoding unit 121 simultaneously decodes the absolute values of a predetermined number of quantized coefficients.

Thus, simultaneous decoding of the absolute values of a predetermined number of quantization coefficients enables simultaneous (parallel) execution of a plurality of processes, and enables decoding of images at higher speed.

Note that although it has been described that the decoding of each of the codes indicating the symbols of w quantization coefficients is sequentially performed for the processing in step S357, an arrangement may be used in which the decoding of Decoding of each of the codes indicating the signs of the w quantized coefficients.

As described above, unlike the case of encoding (or decoding) an image with conventional JPEG 2000, it is not necessary to perform arithmetic encoding of quantization coefficients in increments of bit-planes based on multiple encoding paths, and thus it is possible to perform more efficient processing by simpler processing. Image encoding (or decoding) is performed quickly.

With the conventional JPEG 2000 method, processing is performed for each bit-plane based on multiple encoding paths, therefore, in cases where its processing is performed, quantization coefficients must be accessed approximately as many as obtained by multiplying the quantization coefficients by the number of bit-planes The same number of times, which means a huge amount of processing.

Also, in the case of packetizing already-encoded images, the packetization process cannot be started unless the encoding of one image is completely completed, so a delay equivalent to the waiting interval is generated. Furthermore, by the JPEG 2000 method, the (encoded) quantization coefficients corresponding to the positions within the rectangular area formed by the parallel sides in the x direction and the y direction on the subband shown in FIG. 2, for example, are stored in a Within the group, a delay corresponding to the length of the rectangular area in the y direction is therefore also generated. With the regular JPEG 2000 method, this delay caused by the generative encoding, so real-time processing is difficult. Now, although the delay can be reduced by shortening the length in the y direction of the rectangular area on the subband, the encoding efficiency deteriorates in this case.

In contrast, with the image coding device 11, it is not necessary to perform arithmetic coding of quantized coefficients for each bit-plane based on a plurality of coding paths as described above, and only when coding an image, when a code indicating the absolute value of the quantized coefficient is output , when a code indicating the maximum number of significant digits is output, and when a code indicating the sign of the quantization coefficient is output, the quantization coefficient is accessed, and thus an image can be encoded in a simpler manner.

And, there is a case where the code indicating the maximum significant digit and the code indicating the sign of the quantization coefficient are 1 bit or 0 bit, so in the case of encoding this image, it is possible to access the quantization coefficient only about twice by to actually encode the image. Also, in the case of decoding an image, the quantization coefficient only needs to be accessed once, so the image can be decoded in a simpler and faster manner.

At the image encoding device 11 and the image decoding device 111, the quantized coefficients of the subbands are encoded and decoded in raster scan order, so there is no need to buffer the quantized coefficients, so that delays caused by encoding and decoding can be reduced.

Furthermore, the actual encoding and decoding of a YUV 4:2:2 format image of 1920 pixels horizontally by 1080 vertical pixels using SIMD operation instructions (where w=4) yielded the following results. Note that at the time of performing encoding, an image is wavelet-transformed and decomposed into five sub-band levels, and quantized coefficients obtained by quantizing wavelet coefficients of each sub-band are encoded. And, by causing a CPU (clock frequency of 3.0 GHz) called Pentium (registered trademark) 4 (trademark of Intel Corporation) to execute a predetermined program, functional blocks necessary for encoding and decoding (for example, entropy encoding in FIG. 15 ) are realized. Unit 23 and the entropy decoding unit 121 in Figure 9) and functional blocks for encoding and decoding images with JPEG 2000.

In the case of encoding one frame of an image by the conventional JPEG 2000 method, the code size is 191571 bytes, and the amount of time required for encoding is 0.26157 seconds. And, the time necessary to decode the encoded image is 0.24718 seconds.

In contrast, in the case of encoding one frame of an image by the entropy encoding unit 23 in FIG. 15 , the code size is 343840 bytes, and the amount of time required for encoding is 0.03453 seconds. And, the time necessary to decode the encoded image by the entropy decoding unit 121 in FIG. 9 is 0.02750 seconds.

Moving images are often displayed at 30 frames per second, so as long as encoding or decoding can be performed at 0.033 (=1/30) seconds per frame, images can be processed in real time. By the JPEG 2000 method, the amount of time necessary for encoding is 0.26157 seconds, and the amount of time necessary for decoding is 0.24718 seconds, so real-time processing of images is difficult, but when encoding images with the entropy encoding unit 23 in FIG. 15 In the case of , the amount of time required to encode is 0.03453 seconds, so the image can be processed in exactly approximately real time. Also, in the case of decoding an image with the entropy decoding unit 121 in FIG. 9 , the amount of time required for decoding is 0.02750 seconds, so the image can be sufficiently processed in real time.

In the above, an example of applying the process of encoding or decoding every predetermined number of quantized coefficients to the JPEG 2000 method has been described, but this is not limited to the JPEG 2000 method and can be applied to other encoding methods or decoding methods. Also, in the above, an example of encoding or decoding image data has been described, but this is not limited to image data, and can be applied to a case of encoding or decoding audio data or the like, for example. For example, in the case of encoding audio data, a code indicating the maximum number of significant digits of w predetermined values expressed by a code input as audio data, a code indicating an absolute value of a numerical value, and a code indicating a sign of a numerical value is output as encoded audio data.

Referring to other features of the invention, with the encoding method according to the invention, quantized coefficients are encoded losslessly. Therefore, quantizing higher band coefficients with larger step sizes to match human visual characteristics allows for a significant improvement in image quality per generated code volume. Also, reducing the quantization step size used in a particular spatial range allows for improved image quality in that spatial range.

Furthermore, by referring to the encoding method described in this embodiment, the array of the significant digit part of the absolute value is encoded. If we say, VLC encodes and transmits the significand portion of the absolute value, in the case of N significands of the absolute value, a very large VLC table with 2^(N*W) entries is necessary ( Not only the load and processing time for calculation processing are increased, but also the storage capacity necessary for holding the VLC table is increased). In contrast, with the encoding method according to the present invention, it is not necessary to use such a large table (not only the load and processing time for calculation processing can be reduced, but also the necessary storage capacity can be reduced).

Also, it may be considered to use arithmetic coding with a higher compression rate than VLC, but even if for example a compression method such as JPEG 2000 is used, compared with the case of the coding method according to the present invention, it is not significantly improved Compression ratio. That is, by referring to the encoding method described in this embodiment, although the compression rate is high, the encoding process is easy.

With the encoding method according to the present invention, the maximum significant digits of the absolute values of the coefficients of w sets are encoded, so the fact that the significant digits of adjacent coefficients are similar can be used to reduce the size of the generated code.

Also, with the encoding method according to the present invention, differential encoding is performed when encoding the maximum significant digits of the absolute values of the coefficients of w sets, so it is also possible to reduce the The size of the generated code.

Next, a specific example of the wavelet transform unit 21 of the image encoding device 11 and the wavelet inverse transform unit 123 of the image decoding device 111 will be described. An arrangement may be made in which, at this wavelet transform unit 21, a general-purpose unit that performs analysis filtering in the vertical direction of the screen and in the horizontal direction of the screen, for example, on the image data (picture or layer) of the entire screen processing, and performing division of the entire screen into a plurality of subbands, but using a unit that performs wavelet transformation processing in the method described below enables images to be encoded at a higher speed. In the same manner, an arrangement may be made in which, as the wavelet inverse transform unit 123, a general-purpose unit is used, which performs, for example, on the image data (picture or layer) of the entire screen in the vertical direction of the screen and in the horizontal direction of the screen. Synthesizing filter processing and subbands are synthesized, but using a unit that performs wavelet inverse transform processing in the method described below enables an image to be encoded at a higher speed.

Fig. 19 shows a configuration example of an image encoding device applicable to the first embodiment of the present invention. The image encoding device 401 includes a wavelet transform unit 410 , a midway calculation buffer unit 411 , a coefficient rearrangement buffer unit 412 , a coefficient rearrangement unit 413 , a rate control unit 414 , and an entropy encoding unit 415 .

That is, the wavelet transform unit 410, halfway calculation buffer unit 411, coefficient rearrangement buffer unit 412, and coefficient rearrangement buffer unit 413 of the image encoding device 401 shown in FIG. 19 correspond to the image shown in FIG. The wavelet transform unit 21 of the encoding device 11, and the rate control unit 414 and the entropy encoding unit 415 of the image encoding device 401 shown in FIG. 19 correspond to the quantization unit 22 and the entropy encoding unit of the image encoding device 11 shown in FIG. 1 twenty three.

Image data that has been input is temporarily accumulated in the halfway calculation buffer unit 411 . The wavelet transform unit 410 performs wavelet transform on the image data accumulated in the halfway calculation buffer unit 411 . That is, the wavelet transformation unit 410 reads image data from the midway calculation buffer unit 411, and performs filtering processing by the analysis filter to generate coefficient data having low-band components and high-band components, and midway calculates the buffer The generated coefficient data is stored in unit 411 . The wavelet transformation unit 410 has a horizontal analysis filter and a vertical analysis filter, and performs analysis filter processing on image data groups in two directions of the screen horizontal direction and the screen vertical direction. The wavelet transformation unit 410 reads again the low-band component coefficient data stored in the halfway calculation buffer unit 411, performs filtering processing on the read coefficient data with an analysis filter, and further generates the high-band component and the low-band component coefficient data. The generated coefficient data is stored in the midway calculation buffer unit 411 .

When this processing is repeated and the division level reaches a predetermined level, the wavelet transform unit 410 reads coefficient data from the midway calculation buffer unit 411 , and writes the read coefficient data into the coefficient rearrangement buffer unit 412 .

The coefficient rearrangement unit 413 reads the coefficient data written in the coefficient rearrangement buffer unit 412 in a predetermined order, and supplies to the entropy encoding unit 415 . The entropy encoding unit 415 encodes the supplied coefficient data by, for example, an entropy encoding method such as Huffman encoding or arithmetic encoding.

The entropy encoding unit 415 is controlled so as to operate in conjunction with the rate control unit 414 in which the bit rate of output compression-encoded data is an approximately constant value. That is, the rate control unit 414 supplies a control signal to the entropy encoding unit 415 based on the encoded data information from the entropy encoding unit 415, wherein the entropy encoding The bit rate of data compression-encoded by the unit 415 ends the encoding process of the entropy encoding unit 415 . The entropy encoding unit 415 outputs encoded data at the time when the encoding process ends according to the control signal supplied from the rate control unit 414 .

The processing performed at the wavelet transform unit 410 will be described in more detail. First, a general description of wavelet transform will be given. By wavelet transform of image data as roughly shown in FIG. 20 , the process for segmenting image data into spatial frequencies of high frequency bands and low frequency bands is recursively repeated for data of low frequency band spatial frequencies obtained as a result of the division. . Therefore, compression coding can be performed more efficiently by packing the data of the low-band spatial frequency into a smaller area.

Note that FIG. 20 is an example in a case where the division process of the lowest frequency band component region of image data is repeated three times, divided into a low frequency band component region L or a high frequency band component region H, whereby the division level=3. In FIG. 20, "L" and "H" represent low-band components and high-band components, respectively, and the order of "L" and "H" shows frequency bands as a result of the front side being horizontally divided, and as the rear side The frequency bands that are the result of the vertical split. And, the digits before "L" and "H" indicate the division level of its area.

And as can be seen from the example in FIG. 20 , processing is performed in a stepwise manner from the lower right region to the upper left region of the screen, in which low-band components are driven out. That is to say, with the example in FIG. 20, the area divided into four parts, wherein the area 3HH at the lower right of the screen has the least low-band components (including the most high-band components), which is divided into four parts The area at the upper left of the screen is further divided into four, and among these areas, the area at the upper left is further divided into four. The area in the uppermost left corner is 0LL, which has the most low-band components.

Transformation and division of the low-band components are repeatedly performed because the energy of the image is concentrated in the low-band components. This can also proceed from the state of split level = 1 as an example shown in FIG. 21A to the state of split level = 3 shown in FIG. 21B as an example according to the split level, from the child With understanding. For example, the division level of the wavelet transform in Fig. 20 is 3, thus forming 10 subbands.

The wavelet transformation unit 410 generally uses a filter bank configured with a low-band filter and a high-band filter, and performs processing as described above. Note that digital filters generally have an impulse response, that is, filter coefficients of a plurality of tap lengths, and thus need to perform buffering of coefficient data or input image data in advance only for the amount of filter processing to be performed. Also, similar to the case of performing wavelet transformation on a plurality of stages, wavelet transformation coefficients generated in previous stages need to be buffered only as many times as filter processing can be performed.

Next, as a specific example of this wavelet transformation, a method using a 5×3 filter will be described. The method using a 5×3 filter is an excellent method in that the wavelet transformation can be performed with fewer filter taps, and this method can also be used with the JPEG 2000 standard described with reference to conventional techniques.

The impulse response (Z-transform expression) of the 5×3 filter is configured with a low-band filter H ₀ (z) and a high-band filter H ₁ (z), as in the following expression (3) and expression (4 ) shown. From Expression (3) and Expression (4), it can be seen that the low-band filter H ₀ (z) has 5 taps, and that the high-band filter H ₁ (z) has 3 taps.

H ₀ (z)＝(-1+2z ^-1 +6z ^-2 +2z ^-3 -z ^-4 )/8...(3)

H ₁ (z)=(-1+2z ^-1 -z ^-2 )/2...(4)

According to Expression (3) and Expression (4), the coefficients of the low-band component and the high-band component can be directly calculated. Now, by using lifting techniques, the calculations used for filter processing can be reduced. An overview of processing on the analysis filter side using wavelet transform in the case of applying the lifting technique to a 5×3 filter will be given with reference to FIG. 22 .

In FIG. 22 , the uppermost stage section, the middle stage section, and the lowermost stage section each show a pixel column of an input image, a high-band component output, and a low-band component output. For the uppermost stage, it does not need to be limited to the pixel columns of the input image, but can be the coefficients obtained by the previous filtering process. Here, the uppermost stage part is the pixel columns of the input image, where square marks represent even-numbered (starting from 0) pixels or rows, and circular marks represent odd-numbered pixels or rows.

First, as a first stage, high-band component coefficients d _i ¹ are generated from an input pixel column in Expression (5) below.

d _i ¹ ＝d _i ⁰ -1/2(s _i ⁰ +s _i+1 ⁰ )...(5)

Next, as a second stage, using odd-numbered pixels of the input image, low-band component coefficients s _i ¹ are generated from the following Expression (6).

s _i ¹ =s _i ⁰ +1/4(d _i-1 ¹ +d _i ¹ )...(6)

For the analysis filter side, the image data of the input image is thus divided into low-band components and high-band components by filter processing.

An overview of processing on the synthesis filter side for performing wavelet inverse transform that restores coefficients generated by wavelet transform will be given with reference to FIG. 23 . FIG. 23 corresponds to FIG. 22 described above, uses a 5×3 filter, and shows an example of applying the lifting technique. In FIG. 23 , the uppermost stage part shows input coefficients generated by wavelet transformation, where circular marks represent high-band coefficients and square marks represent low-band coefficients.

First, as a first stage, even-numbered coefficients s _i ¹ (starting from 0) are generated from the input low-band component and high-band component coefficients according to the following expression (7).

s _i ⁰ ＝s _i ¹ -1/4(d _i-1 ¹ +d _i ¹ )...(7)

Next, as the second stage, odd ^- numbered coefficients d i 0 are generated from _the even-numbered coefficients s _i ⁰ generated in the above-mentioned first stage and the input coefficients d _i ¹ of the high-band component according to the following expression (8). .

d _i ⁰ ＝d _i ¹ +1/2(s _i ⁰ +s _i+1 ⁰ )...(8)

For the synthesis filter side, the coefficients of the low-band component and the high-band component are synthesized by filter processing as such, and wavelet inverse transform is performed.

The wavelet transform method will be further described. FIG. 24 shows an example of the filtering process of lifting by the 5×3 filter referring to FIG. 22 , which has been performed up to division level=2. With FIG. 24 , the portion shown on the left side of the figure as an analysis filter is a filter on the wavelet transform unit 410 on the image decoding device 401 side. Also, the portion shown on the right side of the figure as a synthesis filter is a filter on a wavelet inverse transform unit on the image decoding device side described later.

By the following description, taking the pixel at the upper left corner of the screen of a display device or the like as a leading pixel, for example, let us say that pixels are scanned from the left end to the right end of the screen to configure one row, and the scanning of the row is from the upper end to the lower end of the screen Execution, thus configured into a screen.

In FIG. 24 , the left end column shows that the image data positioned corresponding to the rows of the original image data are arranged in the vertical direction. That is, the filtering process by the wavelet transform unit 410 is performed by vertically scanning pixels on the screen using a vertical filter. The filter processing of the division level=1 is shown for the first column to the third column from the left end, and the filter processing of the division level=2 is shown in the fourth column to the sixth column. The second column from the left end shows the high-band component output of the image based on the original image data on the left, and the third column from the left end shows the low-frequency band output based on the original image data and the high-band component Component output. Wherein, for the filter processing of the division level=1, the filter processing of the division level=2 is outputted, as shown in the 4th column to the 6th column from the left end.

By filter processing in which division level=1, high-band component coefficient data is calculated based on original image data pixels as first-stage filter processing, and based on the high-band component coefficient data calculated at the first-stage filter processing and the original image data pixel, computes low-band component coefficient data. Filter processing of one example of division level=1 is shown in the first column to the third column on the left side (analysis filter side) of FIG. 24 . The calculated high-band component coefficient data is stored in the coefficient rearrangement buffer unit 412 described in FIG. 19 . And, the calculated low-band component coefficient data is stored in the halfway calculation buffer unit 411 .

In FIG. 24 , the coefficient rearrangement buffer unit 412 is shown as a portion surrounded by a dotted line, and the midway calculation buffer unit 411 is shown as a portion surrounded by a dotted line.

Based on the result of the filter processing of the division level=1 held in the midway calculation buffer unit 411 , the filter processing in which the division level=2 is executed. For the filter processing of division level=2, the coefficient data calculated as the low-band component coefficient in the filter processing of division level=1 is taken as the coefficient data including the low-band component and the high-band component, and is performed with the division level=1 The filtering process is similar to the filtering process. The high-band component coefficient data and low-band component coefficient data calculated by the filter processing of division level=2 are stored in the coefficient rearrangement buffer unit 412 described with reference to FIG. 19 .

At the wavelet transformation unit 410, the above-described filtering processing is performed in each of the horizontal direction and the vertical direction of the screen. For example, first, filter processing of division level=1 is performed in the horizontal direction, and the generated coefficient data of the high-band component and the low-band component are stored in the midway calculation buffer unit 411 . Next, for the coefficient data stored in the midway calculation buffer unit 411 , filter processing of division level=1 is performed in the vertical direction. Through such processing in the horizontal and vertical directions of division level=1, four regions are formed, which are regions HH each formed by coefficient data obtained by further dividing the high-band component into a high-band component and a low-band component and a region HL, and a region LH and a region LL each formed of coefficient data obtained by further dividing the low-band component into a high-band component and a low-band component.

For division level=2, filter processing is performed on the coefficient data of the low-band component generated by division level=1 in each of the horizontal direction and the vertical direction. That is, for division level=2, the area LL formed by division at division level=1 is further divided into four areas, thereby further forming area HH, area HL, area LH, and area LL within area LL.

With the first embodiment, filtering processing by wavelet transformation is performed multiple times in a staged manner, the processing being divided into increments of several lines in the vertical direction of the screen. With the example in FIG. 24 , the first processing as processing from the first line on the screen performs filtering processing for seven lines, and the second processing from the eighth line and subsequent processing is carried out in four lines. Added to perform filtering processing. The number of lines is based on the number of lines necessary to generate the lowest band component equivalent to one line after division into two parts, a high band component and a low band component.

Hereinafter, a set of rows including other subbands required to generate the lowest band component corresponding to one row (subband coefficient data corresponding to the lowest band component of one row) is called a row block (or partition). Here, one line refers to one line corresponding to pixel data or coefficient data formed in one picture or layer, or in each subband, and corresponds to image data before wavelet transformation. That is, the line block (division) refers to a group of pixel data corresponding to the number of lines necessary to generate coefficient data corresponding to one line of the lowest frequency band component subband after the wavelet transform in the original image data before the wavelet transform, Or the coefficient data group of each sub-band obtained by wavelet transform of its pixel data group.

According to FIG. 24 , the coefficient C5 obtained by the filter processing result of division level=2 is calculated based on the coefficient C _a stored in the midway calculation buffer unit 411, and the coefficient C4 is calculated based on the coefficient stored in the midway calculation buffer unit 411 C _a , coefficient C _b , and coefficient C _c are calculated. Furthermore, the coefficient C _c is calculated based on the coefficient C2 and the coefficient C3 stored in the coefficient rearrangement buffer unit 412 and based on the pixel data of the fifth row. And, the coefficient C3 is calculated based on the pixel data in the fifth row to the seventh row. Thus, in order to obtain the low-band component coefficient C5 of division level=2, the pixel data in the first row to the seventh row are necessary.

On the other hand, for the second and subsequent filter processing, coefficient data that has been calculated and stored in the coefficient rearrangement buffer unit 412 up to the last filter processing can be used, so the number of necessary lines is smaller.

In other words, according to FIG. 24 , among the low-band component coefficients obtained as a result of the filter processing of division level=2, the coefficient C9 which is the next coefficient after the coefficient C5 is based on the coefficient C4 and the coefficient C8, and based on the coefficient stored in the midway calculation buffer unit The coefficient C _c in 411 is calculated. The coefficient C4 has been calculated through the above-described first-time filtering process, and is stored in the coefficient rearrangement buffer unit 412 . Similarly, the coefficient C _c has been calculated through the above-mentioned first-time filtering process, and is stored in the midway calculation buffer unit 411 . Therefore, for this second filtering process, only the filtering process for calculating the coefficient C8 is newly performed. This new filtering process is also performed using the eighth to eleventh lines.

Thus, for the second and subsequent filter processing, data that has been calculated and stored in the midway calculation buffer unit 411 and the coefficient rearrangement buffer unit 412 up to the last filter processing can be used, so it is possible to use only Behavior increments perform processing.

Note that in the case where the number of lines on the screen does not match the number of encoded lines, the lines of the original image data are copied in a predetermined manner so as to match the number of encoded lines, after which filtering processing is performed.

Although the details will be described later, with the present invention, by dividing the lines of the entire screen into several times (increased by line blocks), the filtering process of obtaining only the coefficient data of one line equivalent to the lowest frequency band component is performed in stages, thereby Decoded images can be obtained with minimal delay in the case of transmitted encoded data.

In order to perform wavelet transformation, a first buffer for performing wavelet transformation itself and a second buffer for storing coefficients generated during execution of processing up to a predetermined division level are required. The first buffer corresponds to the midway calculation buffer unit 411, and is shown surrounded by a dotted line in FIG. 24 . Also, the second buffer corresponds to the coefficient rearrangement buffer unit 412 and is shown surrounded by a dotted line in FIG. 24 . The coefficients stored in the second buffer are used in the case of decoding, and thus are objects of entropy encoding processing in the subsequent steps.

Processing at the coefficient rearranging unit 413 will be described. As described above, the coefficient data calculated at the wavelet transform unit 410 is stored in the coefficient rearrangement buffer unit 412 , rearranged and read out by the coefficient rearrangement unit 413 , and transmitted to the entropy encoding unit 415 .

As described above, by wavelet transform, coefficients are generated from the high-band component side to the low-band component side. In the example in FIG. 24 , at the first time, the high-band component coefficient C1 , coefficient C2 , and coefficient C3 are sequentially generated from the pixel data of the original image in the filter processing of division level=1. Then, for the low-band component coefficient data obtained in the filter processing of the division level=1, the filter processing of the division level=2 is performed, thereby sequentially generating the low-band component coefficient C4 and the coefficient C5. That is, for the first time, coefficient data is generated in the order of coefficient C1, coefficient C2, coefficient C3, coefficient C4, and coefficient C5. Based on the principle of wavelet transform, the generation order of coefficient data is always in this order (order from high frequency band to low frequency band).

In contrast, on the decoding side, in order to decode immediately with low delay, it is necessary to generate and output images from low-band components. Therefore, it is desirable to rearrange the coefficient data generated on the encoding side from the low-band component side to the high-band component side and supply it to the decoding side.

A more detailed description will be given with reference to FIG. 24 . The right side of FIG. 24 shows the side of the synthesis filter that performs inverse wavelet transform. Using the lowest frequency band component coefficients C4 and C5 generated at the first filtering process on the encoding side, and the coefficient C1, the first synthesizing process (inverse wavelet transform process) including the output image data of the first line on the decoding side is performed.

That is, through the first synthesis processing, coefficient data is supplied from the encoding side to the decoding side in the order of coefficient C5, coefficient C4, and coefficient C1, so that on the decoding side, by the synthesis processing as corresponding to division level=2 In the processing of synthesis level=2 of , the synthesis processing for the coefficients C5 and C4 is performed to generate the coefficient Cf, and the coefficient Cf is stored in the buffer. Next, synthesis processing for the coefficient Cf and the coefficient C1 is performed by processing of synthesis level=1 which is synthesis processing corresponding to division level=1, thereby outputting the first line.

Thus, by the first synthesis process, the coefficient data generated in the order of coefficient C1, coefficient C2, coefficient C3, coefficient C4, and coefficient C5 on the encoding side and stored in the coefficient rearrangement buffer unit 412 is rearranged as The order of coefficient C5, coefficient C4, coefficient C1, etc., is provided to the decoding side.

Note that with the synthesis filter side shown on the right side of FIG. 24 , the coefficients supplied from the encoding side are referred to by the numbers of the coefficients on the encoding side in parentheses, and the row numbers of the synthesis filters outside the parentheses are shown. For example, coefficient C1(5) shows that it is coefficient C5 on the analysis filter side on the left side of FIG. 24 and is in the first row on the synthesis filter side.

Synthesis on the decoding side by coefficient data generated by the second or subsequent filter processing on the encoding side may be performed using synthesis from the case of the previous synthesis processing, or coefficient data supplied from the encoding side deal with. In the example of FIG. 24 , the second synthesis processing on the decoding side performed using the low-band component coefficient C8 and coefficient C9 generated by the second filtering processing on the encoding side also requires the first filtering processing on the encoding side. The coefficients C2 and C3 are generated, and the second to fifth lines are decoded.

That is, through the second synthesis process, coefficient data is supplied from the encoding side to the decoding side in the order of coefficient C9, coefficient C8, coefficient C2, coefficient C3. On the decoding side, the coefficient _Cg is generated using the coefficients C8 and C9, and the coefficient C4 supplied from the encoding side at the time of the first synthesis process, by the processing of synthesis level=2. Using the coefficient C _g and the above-mentioned coefficient C4 , and C _f generated by the first synthesis process and stored in the buffer, the coefficient _{C h} is generated and stored in the buffer.

By the processing of synthesis level= ₁ , the coefficient _C2 (shown as Coefficient C6(2)), and coefficient C3 (shown as coefficient C7(3) for the synthesis filter) to perform synthesis processing, and decode the second to fifth lines.

Thus, by the second synthesizing process, the coefficient data generated on the encoding side are rearranged as coefficient C2, coefficient C3, (coefficient C4, coefficient C5), coefficient C6, coefficient C7, coefficient C8, coefficient C9, and are rearranged in the form of coefficient The order of C9, coefficient C8, coefficient C2, coefficient C3, etc. is provided to the decoding side.

Thus, for the synthesis processing of the third time and thereafter, similarly, the coefficient data stored in the rearrangement buffer unit 12 is rearranged in a predetermined manner, and supplied to the decoding unit, in which each row is decoded in increments of four rows .

Note that for the synthesis processing on the decoding side corresponding to the filtering processing including the line at the bottom end of the screen on the encoding side, the coefficient data generated up to this point and stored in the buffer are all output, so output The number of rows increases. For the example in Figure 24, eight lines are output during the last pass.

Note that the rearrangement processing of the coefficient data by the coefficient rearrangement unit 413 sets, for example, readout addresses in the case of reading out the coefficient data stored in the coefficient rearrangement buffer unit 412 in a predetermined order.

The above processing will be described in more detail with reference to FIG. 25 . FIG. 25 is an example of performing filtering processing by wavelet transform using a 5×3 filter up to division level=2. By the wavelet transform unit 410, as an example shown in A of FIG. 25 , in each of the horizontal and vertical directions, the first filtering process is performed on the first line to the seventh line of the input image data (FIG. 25 In-1 of A in).

By the processing of division level=1 of the first filter processing, coefficient data of coefficient C1, coefficient C2, and coefficient C3 equivalent to three lines are generated, and as an example shown in B of FIG. 25 , these coefficient data are each Arranged in the region HH, region HL, and region LH formed by the processing of division level=1 (WT-1 in B in FIG. 25 ).

And, the region LL formed by the division level=1 is further divided into four by the filter processing in the horizontal and vertical directions of the division level=2. For the coefficient C5 and the coefficient C4 generated by the division level=2, within the area LL by the division level=1, one line by the coefficient C5 is placed in the area LL, and one line by the coefficient C4 is placed in the area HH, the area HL, and in each of the regions LH.

For the second and subsequent filter processing by the wavelet transform unit 410, the filter processing is performed in increments of four lines (In-2... of A in FIG. 25 ), and coefficient data is generated in increments of two lines at division level=1 (BW-2 in B of FIG. 25 ), and the generation coefficient data is increased by one row at division level=2.

For the example of the second time in FIG. 24 , coefficient data of coefficient C6 and coefficient C7 equivalent to two lines are generated at the filter processing of division level=1, and as an example shown in B of FIG. 25 , these coefficients The data is placed after the coefficient data generated at the first filtering process of the region HH, the region HL, and the region LH formed by the division level=1. In the region LL by the division level=1, the coefficient C9 corresponding to one row generated by the filtering process of the division level=1 is placed in the region LL, and the coefficient C8 equivalent to one row is placed in the regions HH, HL , and in each of the regions LH.

As an example shown in C of FIG. 25, in the case of decoding data subjected to wavelet transformation as in B of FIG. , the first line through the first synthesis process on the decoding side is output (Out-1 in C of FIG. 25 ). Thereafter, four lines are output at a time to the decoding side (Out-2, . Corresponding to the last filtering process on the encoding side, eight lines are output on the decoding side.

The coefficient data generated by the wavelet transform unit 410 from the high-band component side to the low-band component side are sequentially stored in the coefficient rearrangement buffer unit 412 . By the coefficient rearrangement unit 413, when the coefficient data is accumulated in the coefficient rearrangement buffer unit 412 until the above-mentioned coefficient rearrangement can be performed, the coefficient data is rearranged in a necessary order, and is transferred from the coefficient rearrangement buffer unit 412 read out. The read-out coefficient data is sequentially supplied to the entropy encoding unit 415 .

The entropy encoding unit 415 controls the encoding operation so that the bit rate of the output data becomes the target bit rate based on the control signal supplied from the rate control unit 414, and performs entropy encoding of the supplied coefficient data, as with reference to FIGS. 1 to 18 as described. The encoded data subjected to entropy encoding is supplied to the decoding side.

Note that the entropy encoding unit 415 is thus able to perform encoding processing on the supplied coefficient data without performing quantization, but for the case where the coefficient data is first quantized in the same manner as that performed by the quantization unit 22 described with reference to FIG. Arrangements in which the obtained quantization details are subjected to the same source encoding process as that performed by the entropy encoding unit 23 can expect even further improved compression advantages. For the method of this quantization, any method can be used, and for example, a commonly used means for dividing the coefficient data W by the quantization step size Δ, such as shown in the following expression (9) can be used .

Quantization coefficient = W/Δ ... (9)

As described with reference to FIGS. 24 and 25 , with the first embodiment according to the present invention, the wavelet transform unit 410 performs wavelet transform processing in increments of multi-line image data (in increments in line blocks). The encoded data encoded by the entropy encoding unit 415 is output in increments of these line blocks. That is, in the case of performing processing up to division level=2 using a 5×3 filter, for the output of one screen of data, output is obtained, one line for the first time, and four lines each time from the second time to the last time , and output eight lines one last time.

Note that, in the case of performing entropy encoding on the coefficient data after rearrangement by the coefficient rearranging unit 413, for example, in performing entropy encoding on the row of the first coefficient C5 with the first filtering process shown in FIG. 24, for example, case, there are no historical rows, i.e. no rows that have been generated by coefficient data. So in this case only that row is entropy encoded. On the contrary, in the case of encoding the row of the coefficient C1, the row of the coefficient C5 and the coefficient C4 become the history row. A plurality of rows that are close to each other can be considered to be configured with similar data, and thus it is effective to entropy-encode the plurality of rows together.

Also, as described above, with the wavelet transform unit 410 , an example for performing filtering processing by wavelet transform using a 5×3 filter was described, but it should not be limited to this example. For example, for the wavelet transform unit 410, a filter with a longer number of taps, such as a 9×7 filter, may be used. In this case, if the number of taps is longer, the number of lines accumulated in the filter also increases, so the delay time from input of image data until output of encoded data becomes longer.

Also, for the above description, for convenience of description, the division level of wavelet transform is described as division level=2, but it should not be limited thereto, and the division level can be further increased. The more the division level is increased, the better the high compression ratio can be achieved. For example, generally, for wavelet transform, filtering processing up to division level=4 is repeated. Note that the latency increases significantly as the segmentation level increases.

Therefore, when applied to an actual system, it is desirable to determine the number of filter taps or the division level according to the delay time or picture quality of the decoded image required by the system. The number of filter taps or the division level does not need to be a fixed value, and may be appropriately selected.

Next, a specific flow example of the overall encoding process, for example, according to the image encoding device 401 as described above will be described with reference to the flowchart in FIG. 26 .

At the start of the encoding process, in step S401, the wavelet transformation unit 410 sets the number A of the line block to be processed as an initial setting. In general, the number A is set to "1". At the end of the setting, in step S402, the wavelet transformation unit 410 obtains the image data of the number of lines (i.e., one line block) necessary to generate one line of the A-th line from the top of the lowest frequency band subband, for which image data , performing vertical analysis filtering processing in step S403 for performing analysis filtering on image data arranged in the vertical direction of the screen, and performing horizontal analysis filtering processing in step S404 for performing analysis filtering on image data arranged in the horizontal direction of the screen Perform analysis filtering.

In step S405, the wavelet transformation unit 410 determines whether the analysis filter processing has been performed to the final level, and in the case where it is determined that the division level has not reached the final level, the process returns to step S403, where the steps in step S403 and step S404 The analysis filtering process is repeated for the current segmentation level.

In a case where it is determined in step S405 that the analysis filtering process has been performed to the final level, the wavelet transform unit 410 advances the process to step S406.

In step S406, the coefficient rearranging unit 413 rearranges the row block A (the A-th line block from the top of the picture (layer in the case of the interlaced method)) in order from the low frequency band to the high frequency band. coefficient. In step S407, the entropy encoding unit 415 entropy-encodes the coefficients in a row-increasing manner. At the end of the entropy encoding, the entropy encoding unit 415 externally transmits the encoded data of the line block A in step S408.

The wavelet transformation unit 410 increments the value of number A by "one" in step S409, subjects the next line block to processing, and determines whether there is an unprocessed image input line in the picture to be processed in step S410. In a case where it is determined that there is an unprocessed image input line, the process returns to step S402, and the subsequent processes are repeated for a new line block to be processed.

As described above, the processing in steps S402 to S410 is repeatedly performed to encode each line block. When it is determined in step S410 that there is no unprocessed image input line, the wavelet transform unit 410 ends the encoding process of the picture. For the next picture, a new encoding process starts.

In the case of the conventional wavelet transform method, first, horizontal analysis filtering processing is performed on the entire picture, and then vertical analysis filtering processing is performed on the entire picture. Similar horizontal analysis filter processing and vertical analysis filter processing are then sequentially performed on the entire obtained low-band components. As described above, the analysis filtering process is repeated recursively until the segmentation process reaches the final level. Therefore, the result of each analysis filtering process needs to be held in the buffer, but in this case, the buffer needs to hold the filtering result of the entire picture or the entire low-frequency band component of the division level at this point in time, which requires a large Storage capacity (the amount of data to be kept is large).

Also, in this case, if wavelet transformation for the entire picture does not end, coefficient rearrangement or entropy encoding in the subsequent stage cannot be performed, thus significantly increasing delay time.

On the contrary, in the case of the wavelet transform unit 410 of the image encoding device 401, the vertical analysis filter processing and the horizontal analysis filter processing are continuously performed in increments of line blocks up to the final level, as described above, and therefore, one ( During the same period of time) the amount of data held (buffered) is small, thus significantly reducing the storage capacity to be prepared in the buffer. Also, by performing analysis filtering processing up to the last level, later steps for coefficient rearrangement or entropy encoding processing can also be performed (ie, coefficient rearrangement or entropy encoding can be performed in increments of line blocks). Therefore, the delay time can be significantly reduced compared with conventional methods.

FIG. 27 shows one example of the configuration of an image decoding device corresponding to the image encoding device 401 in FIG. 19 . The encoded data (encoded data output in FIG. 19 ) outputted from the entropy encoding unit 415 of the image encoding device 401 in FIG. 19 is supplied to the entropy decoding unit 421 (in FIG. 27 ) of the image decoding device 420 in FIG. coded data input), the entropy code is decoded, and becomes coefficient data. The coefficient data is stored in the coefficient buffer unit 422 . The wavelet inverse transform unit 423 uses the coefficient data stored in the coefficient buffer unit 422, for example, to perform synthesis filter processing as described with reference to FIGS. The result of the filtering process. The wavelet inverse transform unit 423 repeats this process according to the division level to obtain decoded image data (output image data).

The entropy decoding unit 421 of the image decoding device 420 shown in FIG. 27 corresponds to the entropy decoding unit 121 and the inverse quantization unit 122 of the image decoding unit 111 in FIG. 8, and the coefficient buffer unit 422 and the wavelet inverse transform of the image decoding device 420 The unit 423 corresponds to the wavelet inverse transform unit 123 of the image decoding unit 111 .

That is, the description of inverse quantization processing (the inverse quantization unit 122 ), which has little relation to the wavelet inverse transform method, is omitted because the description here is about wavelet inverse transform. Of course, in the case where quantization is to be performed at the image encoding unit 401, an arrangement may be made in which the same inverse quantization as that performed by the inverse quantization unit 122 (inverse quantization corresponding to quantization at the image encoding device 401) may be performed. ), and the obtained wavelet coefficients are supplied to the coefficient buffer unit 422 (wavelet inverse transform unit 123). Thus, further improvement in the compression effect can be expected.

Next, a specific flow example of the entire decoding process by the image decoding device 420 such as described above will be described with reference to the flowchart in FIG. 28 .

At the start of the decoding process, in step S431, the entropy decoding unit 421 obtains encoded data, and in step S432 entropy-decodes the encoded data with behavior increasing. In step S433, the coefficient buffer unit 422 holds the thus decoded and obtained coefficients. In step S434, the wavelet inverse transform unit 423 determines whether coefficients equivalent to one line block have been accumulated in the coefficient buffer unit 422, and if it is determined not to be accumulated, the process returns to step S431, the subsequent processes are performed, and the wavelet The inverse transform unit 423 waits until coefficients equivalent to one line block have been accumulated in the coefficient buffer unit 422 .

In a case where it is determined in step S434 that coefficients equivalent to one line block have been accumulated in the coefficient buffer unit 422, the wavelet inverse transform unit 423 advances the process to step S435, and reads out the coefficients held in the coefficient buffer unit 422. Equivalent to the coefficient of a row block.

The wavelet inverse transform unit 423 performs vertical synthesis filter processing on the read coefficients in step S436, which performs synthesis filter processing on coefficients arranged in the vertical direction of the screen, and in step S437, performs horizontal synthesis filter processing on The coefficients arranged in the horizontal direction of the screen perform synthesis filter processing, and in step S438 it is determined whether the synthesis filter processing has passed level one (a level in which the value of the division level is "1"), that is, it is determined whether inverse conversion has been performed to The state before the wavelet transformation, and if it is determined that level one has not been reached, the process returns to step S436, thereby repeating the filtering process in steps S436 and S437.

In step S438, if it is determined that the inverse transform processing has ended by level 1, the wavelet inverse transform unit 423 advances the process to step S439, and externally outputs the image data obtained by the inverse transform processing.

In step S440, the entropy decoding unit 421 determines whether to end the decoding process, and when it is determined that the input of encoded data is intermittent and the decoding process will not end, the process returns to step S431, and repeats the subsequent steps. deal with. And, in step S440, when the input of the encoded data ends, and so on, so that the decoding process ends, the entropy decoding unit 421 ends the decoding process.

In the case of using the conventional wavelet inverse transform method, first, horizontal synthesis filtering is performed on all coefficients of the division level to be processed in the screen horizontal direction, and then vertical synthesis filtering is performed in the screen vertical direction. That is to say, for each execution of the synthesis filtering process, the result of the synthesis filtering process needs to be held in the buffer, but in this case, the buffer needs to hold the synthesis filtering result of the division level at that time point, and the following All coefficients of a division level, which requires a large storage capacity (the amount of data to be held is large).

Also, in this case, image data output is not performed until all wavelet inverse transformations within the picture (layer in the case of the interlaced method) are completed, so the delay time from input to output increases significantly.

In contrast, in the case of the wavelet inverse transform unit 423 of the image decoding device 420 as described above, the vertical synthesis filter processing and the horizontal synthesis filter processing are continuously performed in increments of line blocks up to level 1, so compared with the conventional method, The amount of data that needs to be buffered at one time (during the same period of time) is small, thus facilitating reduction of the storage capacity to be prepared in the buffer. Also, by performing synthesis filter processing (wavelet inverse transform processing) up to level 1, image data can be sequentially output (increasing by line block) before all image data within a picture is obtained, so delay can be significantly reduced compared to conventional methods time.

Operations of various elements of the image encoding device 401 shown in FIG. 19 or the image decoding device 420 shown in FIG. 27 (the encoding process in FIG. 26 or the decoding process in FIG. 28 ) are not shown, for example, according to a predetermined program. Out of the CPU (Central Processing Unit) control. This program is stored in advance, for example, in an unshown ROM (Read Only Memory). However, this is not restrictive, and the entire device can be operated by interacting with timing signals or control signals between each element including the image encoding device or the image decoding device. Also, the image encoding device or the image decoding device can be realized by software running on a computer device.

Another example of such image encoding processing or image decoding processing will be described next. For example, an arrangement may be made in which various elements of the image encoding device 401 and the image decoding device 420 are concurrently operated in the above-mentioned system (the image encoding device 401 and the image decoding device 420) and thus performed with less delay Image compression encoding and decoding processing. For brevity of description, descriptions of redundant parts of the above contents will be omitted.

FIG. 29 is a schematic view of an example of concurrent operations of various elements of the image encoding device 401 and the image decoding device 420 according to the second embodiment of the present invention. FIG. 29 corresponds to FIG. 25 described above. The first wavelet transform WT-1 (B in FIG. 29 ) is performed on the image data input In-1 (A in FIG. 29 ) at the entropy encoding unit 415 . As described with reference to FIG. 24 , the first wavelet transform WT-1 is started at the time point when the first three lines are input, and the coefficient C1 is generated. That is, from the input of the image data In-1 until the wavelet transform WT-1, a delay equivalent to three lines is generated.

The generated coefficient data is stored in the coefficient rearrangement buffer unit 412 . Thereafter, the input image data is subjected to wavelet transformation, and when the first processing ends, the processing moves to the second wavelet transformation WT-2.

The input of the image data In-2 for the second wavelet transform WT-2 and the rearrangement Ord-1 of the three coefficients are performed concurrently by the coefficient rearrangement unit 413 for the second wavelet transform WT-2 (C of FIG. 29 ). , the three coefficients are coefficient C1, coefficient C4, and coefficient C5.

Note that the delay from the end of wavelet transform WT-1 to the start of rearrangement Ord-1 may be a delay based on device or system configuration, for example, a delay accompanying transmission of a control signal instructing rearrangement processing to coefficient rearrangement unit 413, The delay necessary for the start of processing of the control signal by the coefficient rearranging unit 413 , or the delay necessary for program processing, is not the delay of encoding processing.

The coefficient data is read from the coefficient rearrangement buffer unit 412 in the order in which rearrangement ends, supplied to the entropy encoding unit 415, and subjected to entropy encoding EC-1 (D of FIG. 29 ). The entropy encoding EC-1 can start without waiting for the rearrangement of the three coefficients, namely, the coefficient C1, the coefficient C4, and the coefficient C5, to all end. For example, at a point in time when rearrangement ends for one line from the first output coefficient C5, entropy encoding may be started on the coefficient C5. In this case, the delay from the start of the rearrangement Ord-1 process to the start of the entropy encoding EC-1 process is equivalent to one line.

The encoded data in which the entropy encoding EC-1 by the entropy encoding unit 415 ends is transmitted to the image decoding device 420 via a certain transmission path (E in FIG. 29 ). As a transmission path for transmitting coded data, a communication network such as the Internet can be considered, for example. In this case, the encoded data is transmitted via IP (Internet Protocol). However, the arrangement is not limited thereto, and other transmission paths for encoded data may be communication interfaces such as USB (Universal Serial Bus) and IEEE 1394 (Institute of Electronics and Electrical Engineers 1394), or a communication interface specified by the IEEE 802.11 standard Represents wireless communication, etc.

After the image data equivalent to seven lines is input to the image encoding device 401 by the first processing, the image data down to the lower edge of the screen is sequentially input. With the image encoding device 401, according to image data input In-n (n is 2 or more), as described above, wavelet transform WT-n, rearrangement Ord-n, and entropy encoding EC-n are incrementally performed in four rows. The rearrangement Ord and entropy encoding EC of the last processing at the image encoding device 401 are performed for six lines. These processes are executed concurrently at the image encoding device 401 as in the examples shown in A to D of FIG. 29 .

The encoded data encoded by the entropy encoding EC-1 by the image encoding device 401 is transmitted to the image decoding device 420 via the transmission path, and supplied to the entropy decoding unit 421 . The entropy decoding unit 421 sequentially subjects the supplied encoded data encoded by the entropy encoding EC-1 to entropy-encoded decoding iEC-1, and restores coefficient data (F in FIG. 29 ). The restored coefficient data is sequentially stored in the coefficient buffer unit 422 . When storing only an amount of coefficient data capable of wavelet inverse transform in the coefficient buffer unit 422, the wavelet inverse transform unit 423 reads the coefficient data from the coefficient buffer unit 422, and performs wavelet inverse transform using the read coefficient data iWT-1 (G in Figure 29).

As described with reference to FIG. 24 , the wavelet inverse transform iWT-1 by the wavelet inverse transform unit 423 may start at the time point when the coefficient C4 and the coefficient C5 are stored in the coefficient buffer unit 422 . Accordingly, the delay from the start of the decoding iEC-1 by the entropy decoding unit 421 to the start of the wavelet inverse transform iWT-1 by the wavelet inverse transform unit 423 is equivalent to two lines.

When the wavelet inverse transform iWT-1 equivalent to three lines by the first wavelet transform ends at the wavelet inverse transform unit 423, output Out-1 of the image data generated by the wavelet inverse transform iWT-1 is performed ( FIG. 29 in H). For the output Out-1, the image data of the first line is output as described with reference to FIGS. 24 and 25 .

For the image decoding device 420, after the input of the encoded coefficient data equivalent to three lines through the first processing at the image encoding device 401, the data obtained by entropy encoding EC-n (n is 2 or more) are sequentially input. Encoded coefficient data. With the image decoding means 420, the entropy decoding iEC-n and the wavelet inverse transform iWT-n are performed on the input coefficient data in increments of four lines, as described above, and the restoration of the image data by the wavelet inverse transform iWT-n is sequentially performed Output Out-n. Entropy decoding iEC and wavelet inverse transform iWT corresponding to the last time of the image encoding device are performed on six lines, and output Out outputs eight lines. These processes are executed concurrently at the image decoding device, as in the examples shown in F to H of FIG. 29 .

As described above, by concurrently performing various processes at the image encoding device 401 and the image decoding device 420 in order from the upper part to the lower part of the screen, image compression processing and image decoding processing can be performed with little delay.

The delay time from image input until image output in the case of performing wavelet transform using a 5×3 filter up to division level=2 is calculated with reference to FIG. 29 . The delay time from the input of the image data of the first line to the image encoding device 401 until the output of the image data of the first line from the image decoding device 420 becomes the sum of the elements listed below. Note that delays that differ due to system configurations, such as delays in transmission paths or delays according to actual processing timings of parts of the device, are excluded.

(1) Delay D_WT from the input of the first line until the end of the wavelet transform WT-1 equivalent to seven lines

(2) According to the time D_Ord of rearranging Ord-1 equivalent to a total of three rows

(3) According to the time D_EC of entropy coding EC-1 equivalent to three lines

(4) The time D_iEC of iEC-1 is decoded according to the entropy equivalent to three lines

(5) According to the time D_iWT of the wavelet inverse transform iWT-1 equivalent to three lines

Delays from the above elements will be calculated with reference to FIG. 29 . The delay D_WT in (1) corresponds to the time of ten lines. Time D_Ord in (2), time D_EC in (3),

Each of the time D_iEC in (4) and the time D_iWT in (5) corresponds to the time of three lines. On the other hand, with the image encoding device 401 , entropy encoding EC-1 may start one line after rearrangement Ord-1. Similarly, for the image decoding device 420, the inverse wavelet transform iWT-1 may start two lines after the start of the entropy decoding iEC-1. Also, the entropy decoding iEC-1 may start processing at the point in time when encoding equivalent to one line at the entropy encoding EC-1 is completed.

Therefore, with the example in FIG. 29 , the delay time from when the image data of the first line is input to the image encoding device until the image data of the first line is output from the image decoding device is equivalent to 10+1+1+2+3=17 lines.

A more specific example considering the delay time will be given. In the case where the input image data is image data of an interlaced video signal for HDTV (High Definition Television), one frame is configured with a resolution of 1920 pixels by 1080 lines, for example, and one layer is 1920 pixels by 1080 lines. Take 540 lines. Therefore, if the frame frequency is 30 Hz, one field of 540 lines is input to the image encoding device 401 in a time of 16.67 milliseconds (=1 second/60 layers).

Therefore, the delay time from input of image data equivalent to seven lines is 0.216 milliseconds (=16.67 milliseconds×7/540 lines), which is an extremely short time for an update time of one layer, for example. And, regarding the sum of delay D_WT in (1), time D_Ord in (2), time D_EC in (3), time D_iEC in (4), and time D_iWT in (5) above, to be processed The number of rows is relatively small, so the delay time is extremely shortened. If elements performing various processes are arranged as hardware, the processing time can be further shortened.

Yet another example will be described. For the description made above, coefficient data is rearranged at the image encoding device 401 after wavelet transformation is performed. In contrast, a case will be described below regarding rearrangement of coefficient data after entropy encoding is performed. With the image encoding device in this case, entropy encoding is performed on coefficients generated by wavelet transforming input image data, and rearrangement processing is performed on entropy-encoded data. Thus, by performing coefficient data rearrangement after performing entropy encoding, the storage capacity required for the coefficient rearrangement buffer can be suppressed.

For example, when the bit precision of the input image data is 8 bits, if wavelet transform is performed up to multiple levels of division, the bit precision of the generated coefficient data becomes, for example, about 12 bits. In the case where the coefficient rearrangement process is performed before the entropy encoding process, the coefficient rearrangement buffer unit needs to store coefficient data with a bit accuracy of 12 bits equivalent to a predetermined number of lines. By arranging coefficient data to be subjected to rearrangement processing generated by wavelet transform after entropy encoding, the coefficient rearrangement buffer can store data compressed by entropy encoding, thus requiring only a small storage capacity.

Fig. 30 shows an example of an image encoding device in this case. Portions in FIG. 30 that are common to those in FIG. 19 described above have the same reference numerals, and descriptions thereof will be omitted.

The input image data is temporarily stored in the midway calculation buffer unit 411 of the image encoding device 430 . The wavelet transform unit 410 performs predetermined wavelet transform on the image data stored in the midway calculation buffer unit 411, as described in the first embodiment. The coefficient data generated by wavelet transform is supplied to the entropy encoding unit 415 . The entropy encoding unit 415 operates in conjunction with the rate control unit 414, and is controlled so that the bit rate of output compression-encoded data becomes a substantially fixed value, and performs entropy encoding processing on the supplied coefficient data. That is, the entropy encoding unit 415 encodes the obtained coefficients in the same order as they were obtained regardless of the order of the coefficients.

Coded data in which coefficient data generated by wavelet transformation has been entropy coded by the entropy coding unit 415 is temporarily stored in the code rearrangement buffer unit 431 . When encoded data to be rearranged is stored in the code rearrangement buffer unit 431 , the code rearrangement unit 432 rearranges the encoded data and reads out the encoded data from the code rearrangement buffer unit 431 . As has been described with reference to the first embodiment, the coefficient data generated by the wavelet transform unit 410 is generated in order from high-band components to low-band components from the upper end side of the screen to the lower end side. In order to output image data with a small delay on the decoding side, the encoded data stored in the encoding rearrangement buffer unit 431 is rearranged in the order from the low-band component to the high-band component of the coefficient data by wavelet transform, and is read out.

The encoded data read out from the encoding rearrangement buffer unit 431 is transmitted to the transmission path as output encoded data, for example.

Note that data encoded and output by the image encoding device 430 can be decoded by the image decoding device 420 that has been described with reference to FIG. 27 in the same manner as when data encoded by the image encoding device 401 is decoded. That is to say, encoded data input into the image decoding device 420 via a transmission path, for example, undergoes entropy-encoded decoding at the entropy decoding unit 421 , and restores coefficient data. The restored coefficient data is sequentially stored in the coefficient buffer unit 422 . The wavelet inverse transform unit 423 performs wavelet inverse transform on the coefficient data stored in the coefficient buffer unit 422, and outputs image data.

Yet another example will be described. The above has been described so that rearrangement processing of coefficient data generated by wavelet transform is performed on the image encoding device side, as shown in an exemplary form in FIG. 31 . Here, a description will be made regarding a case where rearrangement processing of coefficient data generated by wavelet transform is performed on the image decoding device side, as shown in an exemplary form in FIG. 32 .

For rearrangement processing of coefficient data generated by wavelet transform, a relatively large capacity is required as a storage capacity for a coefficient rearrangement buffer, and a high processing capability is also required for the processing itself of the coefficient rearrangement process. Also in this case, if the processing capability on the image encoding device side is higher than a certain amount, even if the coefficient rearrangement processing is performed on the image encoding device side, as described above, no problem occurs.

Here, a case will be considered in which an image encoding device is mounted on a device with relatively low processing capability, such as a mobile terminal such as a cellular phone terminal or a PDA (Personal Digital Assistant). For example, recently, a product in which a camera function is added to a cellular phone terminal (referred to as a cellular phone terminal with camera function) has been widely used. A case may be considered in which image data captured by an image of a cellular phone device having a camera function is compression-encoded by wavelet transformation and entropy encoding, and transmitted via wireless or cable communication.

Such a mobile terminal is limited by its CPU processing capability, and also has an upper limit of storage capacity. Therefore, the load for processing with the above-mentioned coefficient rearrangement is a problem that cannot be ignored.

Thus, as an example shown in FIG. 32, by putting rearrangement processing on the image decoding device side, the load on the image encoding device side can be lightened, thus enabling the image encoding device to be installed in a mobile terminal such as on devices with relatively low processing power.

Fig. 33 shows an example of the configuration of an image encoding device in this case. Note that in FIG. 33 , portions common to those of FIG. 19 described above are given the same reference numerals, and detailed descriptions thereof are omitted.

The configuration of the image coding device 441 shown in FIG. 33 is arranged as a configuration in which the coefficient rearrangement unit 413 and the coefficient rearrangement buffer unit 412 are obtained from the configuration of the image coding device 401 shown in FIG. 19 described above. remove. In other words, with the fourth embodiment, the image encoding device 441 uses a configuration that combines the wavelet transform unit 410 , midway calculation buffer unit 411 , entropy encoding unit 415 , and rate control unit 414 .

Input image data is temporarily accumulated in the midway calculation buffer unit 411 . The wavelet transform unit 410 performs wavelet transform on the image data accumulated in the midway calculation buffer unit 411 , and sequentially supplies the generated coefficient data to the entropy encoding unit 415 in the order of the generated coefficient data. That is, the generated coefficient data is supplied to the entropy encoding unit 415 in the order from the high-band component to the low-band component in the order according to the wavelet transform. The entropy encoding unit 415 performs entropy encoding on the supplied coefficients, and the bit rate of output data is controlled by the rate control unit 414 . The coefficient data generated by the wavelet transform is output as encoded data that has been entropy encoded.

FIG. 34 shows one example of the configuration of an image decoding device in this case. Note that in FIG. 34 , portions common to those in FIG. 27 described above are given the same reference numerals, and detailed descriptions thereof are omitted.

The encoded data output from the entropy encoding unit 415 of the image encoding device 441 described in FIG. 33 is supplied to the entropy decoding unit 421 of the image decoding device 442 in FIG. 34 , entropy encoded, and becomes coefficient data. The coefficient data is stored in the coefficient rearrangement buffer unit 443 via the coefficient buffer unit 442 . When the coefficient data is accumulated in the coefficient rearrangement buffer unit 443 until the coefficient data can be rearranged, the wavelet inverse transform unit 423 pairs the values stored in the coefficient rearrangement buffer unit 443 in the order from the low-band component to the high-band component. The coefficient data is rearranged, and the coefficient data stored in the coefficient rearrangement buffer unit 443 is read out, and then wavelet inverse transform processing is performed using the coefficient data in the order read out. In the case of using a 5×3 filter, its arrangement is as shown in FIG. 32 described above.

That is, for processing from one frame, for example, at a point in time when the coefficient C1, coefficient C4, and coefficient C5 decoded with entropy encoding are stored in the coefficient rearrangement buffer unit 443, the wavelet inverse transform unit 423 reads coefficient data from the coefficient rearrangement buffer unit 443, and performs wavelet inverse transform processing. Data subjected to wavelet inverse transformation by the wavelet inverse transformation unit 423 is sequentially output as output image data.

Note that also in this case, as already described with reference to FIG. 29 , the processing by the various elements of the image encoding device 441 and the transmission of encoded data to the transmission path, and the various elements of the image decoding device 442 can be performed concurrently. Component handling.

Next, description will be made regarding the exchange of encoded data between the above-described image encoding device and image decoding device. Fig. 35 is a diagram for describing an example of how encoded data is exchanged. In the case of the example shown in FIG. 3 , similarly to the other embodiments described above, image data is subjected to wavelet transformation (subband 451 ) while being input in increments of line blocks to correspond to only a predetermined number of lines. . In case a predetermined wavelet transform partition level is reached, the coefficient rows from the lowest frequency band subband to the highest frequency band subband are rearranged in the reverse order to the order in which they were generated, i.e. in the order from low frequency band to high frequency band was rearranged.

For the sub-band 451 in FIG. 35 , the parts divided by the patterns of diagonal lines, vertical lines, and wavy lines are each different row blocks (as indicated by the arrows, the white space in the sub-band 451 is also represented by row blocks. increased and split, and processed). The coefficients of the line blocks after rearrangement are subjected to entropy encoding as described above, thus generating encoded data.

Here, if the image encoding device, for example, transmits the encoded data as it is, it may be difficult for the image decoding device to recognize the boundary of each line block (or require complicated processing). Thus, with the present embodiment, an arrangement is made in which the image encoding apparatus adds a header to encoded data in increments of, for example, line blocks, and transmits a packet formed of the header and the encoded data.

In other words, when the image encoding device generates encoded data (encoder data) of the first line block (Lineblock-1), the encoded data is packetized and transmitted to the image decoding device as a transmission packet 461, as shown in FIG. 35. When the image decoding device receives this packet (received packet 471), its encoded data is decoded.

Similarly, when the image encoding device generates encoded data of the second line block (Lineblock-2), the encoded data is packetized and transmitted to the image decoding device as a transmission packet 462 . When the image decoding device receives this packet (received packet 472), its encoded data is decoded. Still similarly, when the image encoding device generates encoded data of the third line block (Lineblock-3), the encoded data is packetized and transmitted to the image decoding device as a transmission packet 463 . When the image decoding device receives this packet (received packet 473), its encoded data is decoded.

The image encoding device and the image decoding device repeat the above-described processing up to the last X-th line block (Lineblock-X) (transmission packet 464, received packet 474). Thus, a decoded image 481 is generated at the image decoding device.

Fig. 36 shows a configuration example of a header. As described above, the packet includes a header (Header) 491 including a description of the line block number (NUM) 493 and the length of encoded data (LEN) 494, and encoded data.

The image decoding device can quickly identify the boundary of each line block by reading this information included in the header added to the received encoded data, thereby reducing the load of decoding processing or processing time.

Note that, as shown in FIG. 36 , a description may be further added with the subbands constituting the line block being increased quantization step sizes (Δ1 to ΔN) 492 . Thus, the image decoding device can perform inverse quantization in increments of subbands, and thus can perform more detailed image quality control.

Also, the image encoding device and the image decoding device may be arranged to perform the above-described various processes of encoding, packetization, packet switching, and decoding as described with reference to the fourth embodiment concurrently (in a pipelined manner) in increments of line blocks.

Thus, the delay time until image output is obtained at the image decoding device can be significantly reduced. As an example, Fig. 35 shows an operation example for an interlaced motion picture (60 layers/sec). For this example, the time for one layer is 1 second ÷ 60 = about 16.7 milliseconds, but by performing various processes concurrently, image output can be arranged to be obtained with a delay time of about 5 milliseconds.

As described in the above examples, the image encoding device (or image decoding device) can perform encoding processing (or decoding processing) at higher speed by performing wavelet transformation (or wavelet inverse transformation) in increments of line blocks.

That is, the generated code size can be reduced by utilizing the feature of the order in which coefficients are output from a wavelet transform unit that performs transform in increments of line blocks (ie, a feature that the significant digits of consecutive coefficients are similar). Also, even in the case of rearranging coefficients, the wavelet transform unit performs wavelet transform in increments of line blocks, therefore, the feature that the significant digits of consecutive coefficients are similar is not significantly affected, and the generated The code size does not change significantly.

As described above, the above-mentioned wavelet transform processing, and the entropy encoding processing according to the present invention are similar in characteristics and expected advantages of the coefficient data they process, and have a strong affinity with each other. Therefore, for the overall image coding process, more advantages can be expected from applying the above-described wavelet transform processing than applying usual wavelet transform processing.

As described above, such wavelet transform processing can be realized by widely different various configurations, and the entropy encoding processing and entropy decoding processing (image encoding device and image decoding device) according to the present invention can be applied to widely different various systems .

Specific application examples will be described below. First, an example of a digital triax system to which the image encoding device and the image decoding device according to the above-described embodiments have been applied will be described.

A triax system is a system used in television broadcast stations, production studios, and the like. For such systems, a single triaxial cable connecting the video camera to the camera control unit or switch is used to transmit multiplexed signals such as picture signals, audio signal, return picture signal, synchronization signal, etc., and is also used for power supply.

Many conventional three-axis systems have been arranged to transmit the above-mentioned signals in the form of analog signals. On the other hand, in recent years, all systems have become digital, and thus the three-axis system used in television broadcasting stations has also become digital.

With known triaxial systems, the digital video signal transmitted over the triaxial cable has always been an uncompressed video signal. The reason for this is that the specifications required regarding the signal delay time are particularly strict for TV broadcasting stations, and basically, the delay time from shooting to monitor output is required to be within one layer (16.67 milliseconds), for example. Compression encoding systems such as MPEG2 and MPEG4, which have achieved high compression rate and high image quality, have not been used in the three-axis system because video signal compression and encoding, and decoding of the compressed video signal require as much as several frames A lot of time, which means a lot of delay time.

As described above, the image encoding and image decoding methods according to the present invention have an extremely short delay time from inputting image data to obtaining an output image, which is within one layer time, that is, several lines to dozens of lines, so that it is possible to Digital triaxial system for suitable application.

FIG. 37 shows the configuration of an example of a digital triaxial system applicable to the image encoding and image decoding methods according to the present invention. The transmission unit 500 and the camera control unit 502 are connected via a triax cable (triax cable) 501 . A digital video signal and a digital audio signal (referred to as a "main line signal" herein), which are actually broadcast or used as content from the transmission unit 500 to the camera control unit 502 , are transmitted via this triax cable 501 , and from the camera control unit 502 The internal communication of the control unit 502 to the video camera unit 503 is an audio signal and a digital video signal back.

The transmission unit 500 is built in, for example, an unillustrated video camera device. Of course, other arrangements may be made, such as the transmission unit 500 being connected to the video camera device as an external device of the video camera device. The camera control unit 502 may be, for example, a device generally called a CCU (Camera Control Unit).

The digital audio signal has little to do with the essence of the present invention, so its description will be omitted for the sake of brevity.

The video camera unit 503 is configured, for example, in an unshown video camera device, and performs light reception of light from a subject that has passed through a camera including a lens, An optical system 550 such as a focus mechanism, a zoom mechanism, and an aperture adjustment mechanism is captured. The image pickup device converts the received light into an electric signal by photoelectric conversion, and further performs predetermined signal processing for output as a baseband digital video signal. These digital video signals are mapped to, for example, HD-SDI (High-Definition Serial Data Interface) format, and output.

A display unit 551 serving as a monitor and an intercom 552 for exchanging audio externally are also connected to the video camera unit 503 .

The transmission unit 500 has a video signal encoding unit 510 and a video signal decoding unit 511 , a digital modulation unit 512 and a digital demodulation unit 513 , amplifiers 514 and 515 , and a video separation/synthesis unit 516 .

A baseband digital video signal mapped to, for example, HD-SDI format is supplied from the video camera unit 503 to the transmission unit 500 . The digital video signal is compressed and encoded at the video signal encoding unit 510 so as to become a code stream, which is supplied to the digital modulation unit 512 . The digital modulation unit 512 modulates the supplied code stream into a signal in a format suitable for transmission on the triaxial cable 501, and outputs it. The signal output from the digital modulation unit 512 is supplied to the video separation/synthesis unit 516 via the amplifier 514 . The video separation/synthesis unit 516 sends the supplied signal to the triaxial cable 501 . These signals are received at the camera control unit 502 via the triaxial cable 501 .

The signal output from the camera control unit 502 is received at the transmission unit 500 via the triaxial cable 501 . The received signal is supplied to a video separation/synthesis unit 516, and parts of the digital video signal and parts of other signals are separated. Of the received signals, part of the digital video signal is supplied to the digital demodulation unit 513 via the amplifier 515, and the signal modulated into a signal of a format suitable for transmission on the triaxial cable 501 is decoded on the camera control unit 502 side. tuned, and code flow is resumed.

The code stream is supplied to the video signal decoding unit 511, the compression encoding is decoded, and a baseband digital video signal is obtained. The decoded digital video signal is mapped to the HD-SDI format, and output, and supplied to the video camera unit 503 as a returned digital video signal. The returned digital video signal is supplied to a display unit 551 which is connected to the video camera unit 503 and is used for monitoring by the camera operator.

The camera control unit 502 has a video separation/synthesis unit 520 , amplifiers 521 and 522 , a front end unit 523 , a digital demodulation unit 524 and a digital modulation unit 525 , and a video signal decoding unit 526 and a video signal encoding unit 527 .

The signal output from the transmission unit 500 is received at the camera control unit 502 via the triaxial cable 501 . The received signal is provided to the video separation/synthesis unit 520 . The video separation/synthesis unit 520 supplies the signal supplied thereto to the digital demodulation unit 524 via the amplifier 521 and the front end unit 523 . Note that the front-end unit 523 has a gain control unit for adjusting the gain of an input signal, a filter unit for performing predetermined filtering on the input signal, and the like.

The digital demodulation unit 524 demodulates the modulated signal into a signal in a format suitable for transmission on the triaxial cable 501 on the transmission unit 500 side, and restores the code stream. This code stream is supplied to a video signal decoding unit 526, where the compression encoding is decoded to obtain a baseband digital video signal. The decoded digital video signal is mapped to HD-SDI format and output, and externally output as a main line signal.

The returned digital video signal and digital audio signal are externally supplied to the camera control unit 502 . The digital audio signal is provided, for example, to the camera operator's intercom 552 for use in communicating external audio instructions to the camera operator.

The returned digital video signal is supplied to the video signal encoding unit 527 and compression-encoded, and supplied to the digital modulation unit 525 . The digital modulation unit 525 modulates the supplied code stream into a signal in a format suitable for transmission on the triaxial cable 501, and outputs it. The signal output from the digital modulation unit 525 is supplied to the video separation/synthesis unit 520 via the front end unit 523 and the amplifier 522 . The video separation/synthesis unit 520 multiplexes these signals with other signals, and sends to the triaxial cable 501 . These signals are received at the video camera unit 503 via the triaxial cable 501 .

For this seventh embodiment of the present invention, the image encoding device and the image decoding device described by the above embodiments are respectively applied to the video signal encoding unit 510 and the video signal encoding unit 527, and the video signal decoding unit 511 and the video signal decoding unit 526.

In particular, the second embodiment of the present invention arranged so that the processing of various elements at the image encoding device and the image decoding device can be executed in parallel can significantly suppress the output from the camera control unit 502 at the video camera unit 503 The delay of the picture, and the delay of the return digital video signal supplied externally and transmitted from the camera control unit 502 to the video camera unit 503, are suitably applied to the seventh embodiment of the present invention.

Also, in the case of the system shown in FIG. 37 , the signal processing capability and the storage capacity can be appropriately set at each of the transmission unit 500 and the camera control unit 502, so the position where coefficient data rearrangement processing is performed can be is on either the transmission unit 500 side or the camera control unit 502 side, and in the same manner, the location for performing entropy encoding may be before or after rearrangement processing.

That is, on the transmission unit 500 side, the video signal encoding unit 510 performs wavelet transformation and entropy encoding according to the method of the present invention on the digital video signal supplied thereto, and outputs a code stream. As described above, the video signal encoding unit 510 starts wavelet transformation when the number of lines corresponding to the number of filter taps for wavelet transformation and according to the number of division levels of wavelet transformation is input. Further, as described above with reference to FIG. 23 , FIG. 24 , FIG. 29 , etc., processing is sequentially performed by each component while coefficient data necessary for each component is accumulated at the image encoding device and the image decoding device. When processing ends at the bottom row of a frame or layer, processing for the next frame or layer begins.

This is also true for the return digital video signal transmitted from the camera control unit 502 side to the transmission unit 500 side. That is, on the camera control unit 502 side, wavelet transformation and entropy encoding according to the present invention are performed on the returned digital video signal supplied externally by the video signal encoding unit 527, and a code stream is output.

Now, there are many cases where the return digital video signal is allowed to have a lower picture quality than that of the main line signal. In this case, the bit rate at the moment of encoding at the video signal encoding unit 527 can be reduced. For example, the video signal encoding unit 527 performs control through the rate control unit 414 so that the bit rate of the entropy encoding process at the entropy encoding unit 415 is lower. Also, an arrangement can be conceived in which, for example, at the camera control unit 502, transformation processing is performed to a higher division level through the wavelet transformation unit 410 at the video signal encoding unit 527, and at the transmission unit 500 , the wavelet inverse transform at the wavelet inverse transform unit 423 on the video signal encoding unit 511 side stops at a lower division level. Processing at the video signal encoding unit 527 of the camera control unit 502 is not limited to this example, and various other types of processing can be conceived, such as keeping the division level for wavelet transformation low in order to reduce the load of transformation processing .

Next, an example of a case will be described in which an image encoding device and an image decoding device according to the present invention are applied to a system that performs transmission of encoded data using wireless communication. Fig. 38 shows the configuration of an example of a wireless transmission system according to an eighth embodiment of the present invention. Note that in the example in FIG. 38 , video signals are transmitted unidirectionally from the video camera or the transmission unit 600 side (hereinafter will be simply referred to as “transmission unit 600 ”) to the reception apparatus 601 side. Two-way communication between the transmission unit 600 and the reception device 601 can be performed on audio signals and other signals.

The transmission unit 600 is incorporated, for example, in a video camera arrangement, not shown, with a video camera unit 602 . Of course, other arrangements are possible such that the transmission unit 600 is connected to the video camera device as an external device of the video camera device having the video camera unit 602 .

The video camera unit 602 has a predetermined optical system such as an image pickup device such as a CCD, and a signal processing unit for outputting a signal output from the image pickup device as a digital video signal. These digital video signals are mapped to the HD-SDI format, for example, and output from the video camera unit 602, for example. Of course, the digital video signal output from the video camera unit 602 is not limited to this example, and may also be in other formats.

The transmission unit 600 has a video signal encoding unit 610 , a digital modulation unit 611 , and a wireless module unit 612 . At the transmission unit 600 , the baseband digital video signal is mapped to the HD-SDI format, for example, and output from the video camera unit 602 . The digital video signal is compression-encoded at the video signal encoding unit 610 by wavelet transformation and entropy encoding according to the compression encoding method of the present invention to become a code stream, which is supplied to the digital modulation unit 611 . The digital modulation unit 611 performs digital modulation, modulates the supplied code stream into a signal in a format suitable for wireless communication, and outputs it.

And, digital audio signals and other signals, such as predetermined commands and data, for example, are also supplied to the digital modulation unit 611 . For example, the video camera unit 602 has a microphone by which collected sound is converted into an audio signal, which in turn is A/D converted, and output as a digital audio signal. Further, the video camera unit 602 is capable of outputting certain commands and data. Commands and data may be generated within the video camera unit 602, or an operation unit may be provided to the video camera unit 602, wherein commands and data are generated in response to user operations made at the operation unit. Also, an arrangement may be made in which an input device for inputting commands and data is connected to the video camera unit 602 .

The digital modulation unit 611 performs digital modulation on these data audio signals and other signals, and outputs. The digitally modulated signal output from the digital modulation unit 611 is supplied to the wireless module unit 612 and wirelessly transmitted from the antenna 613 as airwaves.

Upon receiving ARQ (Automatic Repeat Request) from the reception device 601 side, the wireless module unit 612 notifies this ARQ to the digital modulation unit 611 in order to request data retransmission.

The electric wave transmitted from the antenna 613 is received at the antenna 620 on the receiving device 601 side, and is supplied to the wireless module unit 621 . The wireless module unit 621 supplies a digitally modulated signal based on the received electric wave to the front end unit 622 . The front-end unit 622 performs predetermined signal processing such as gain control on the supplied digital modulated signal, for example, and supplies it to the digital demodulation unit 623 . The digital demodulation unit 623 demodulates the supplied digitally modulated signal, and restores the code stream.

The code stream restored at the digital demodulation unit 623 is supplied to the video signal decoding unit 624, the compression encoding is decoded by the decoding method according to the present invention, and a baseband digital video signal is obtained. The decoded digital video signal is mapped to, for example, HD-SDI format, and output.

The digital demodulation unit 623 is also supplied with digital audio signals and other signals that are digitally modulated on the transmission unit 600 side and transmitted. The digital demodulation unit 623 demodulates signals in which these digital audio signals and other signals have been digitally modulated, and restores and outputs the digital audio signals and other signals.

And, the front end unit 622 performs error detection according to a predetermined method with respect to the received signal supplied from the wireless module unit 632 , and outputs ARQ in a case where an error is detected such as, for example, that an erroneous frame has been received. ARQ is supplied to the wireless module unit 621 and transmitted from the antenna 620 .

With this configuration, for example, the transmission unit 600 is placed in a relatively small-sized video photography device having a video camera unit 602, a monitor device is connected to the reception device 601, and the digital video signal output from the video signal decoding unit 624 is Provided to the monitor device. As long as the receiving device 601 is within the range of the radio wave transmitted from the wireless module unit 612 of the video camera device with the built-in transmission unit 600, it can be transmitted with a small delay on the monitor device, for example, within one layer or one frame. Latency, viewing pictures taken with video camera setup.

Note that, in the example shown in FIG. 38 , communication between the transmission unit 600 and the reception device 601 is performed using wireless communication to transmit video signals via wireless communication, but the arrangement is not limited to this example. For example, the transmission unit 600 and the reception device 601 may be connected via a network such as the Internet. In this case, the wireless module unit 612 at the transmission unit 600 side and the wireless module unit 612 at the reception device side 601 are each a communication interface capable of communication using IP (Internet Protocol).

Various applications to the system can be conceived from this eighth embodiment. For example, the system according to this eighth embodiment can be applied to a video conference system. One example of arrangement may be to connect a simple video camera device capable of USB (Universal Serial Bus) connection to a computer device such as a personal computer with the computer device side implementing the video signal encoding unit 610 and the video signal decoding unit 624 . The video signal encoding unit 610 and the video signal decoding unit 624 implemented at the computer device may be a hardware configuration, or may be realized by software running on the computer device.

For example, each member participating in a video conference will be provided with a computer device and a video camera device connected to the computer device, and the computer device is connected to a server device for providing video conference system service through a cable or a wireless network. A video signal output from a video camera device is supplied to a computer device via a USB cable, and encoding processing according to the present invention is performed at a video signal encoding unit 610 within the computer device. The computer device transmits the code stream in which the video signal has been encoded to the server device or the like via the network.

The server device transmits the received code stream to the computer device of each participating member via the network. This code stream is received at each participating member's computer device, and the decoding process according to the present invention is performed at the video signal decoding unit 624 within the computer device. The image data output from the video signal decoding unit 624 is displayed on the display unit of the computer device as a picture.

That is, video pictures taken by video camera devices of other participating members are displayed on the display unit of each participating member's computer device. Therefore, for the eighth embodiment of the present invention, the delay time from encoding to decoding the video signal picked up by the video camera device at the computer device of the other participating member is very short, so the display on the displayed participating member The unnatural feeling of pictures of other participating members on the display unit of the computer device can be reduced.

Furthermore, an arrangement is conceivable in which the video signal encoding unit 610 is installed on the video camera device side. For example, the transmission unit 600 is placed in a video camera device. This configuration eliminates the need for a video camera to be connected to another device such as a computer device or the like.

Such a system consisting of a video camera device having a built-in transmission unit 600, and a receiving device 601 can be applied to various applications in addition to the above-mentioned video conference system. For example, as schematically shown in FIG. 39, this system can be applied to a home game machine. In FIG. 39 , a transmission unit 600 according to an eighth embodiment of the present invention is built in a video camera device 700 .

In the main unit 701 of the home game machine, the bus connects, for example, a CPU, RAM, ROM, and a disk drive device compatible with CD-ROM (Compact Disk Read Only Memory) and DVD-ROM (Digital Versatile Disk ROM), for A graphics control unit that converts a display control signal generated by a CPU into a video signal and outputs it, an audio playback unit for playing an audio signal, and the like have a configuration generally similar to a computer device. The main unit 701 of the home game machine is generally controlled by the CPU following the programs stored in the ROM in advance, or the programs recorded in the CD-ROM or DVD-ROM installed in the disk drive. RAM is used as the working memory of the CPU. The main unit 701 of the home game machine incorporates the receiving device 601 . Digital video signals and other signals output from the receiving device 601 are supplied to the CPU via a bus, for example.

We can say that for such a system, the main unit 701 of a home game machine, for example, runs game software capable of ingesting images in the form of externally supplied digital video signals as in-game images. For example, this game software can use images in the form of digital video signals provided externally as images within the game, and also recognize the movement of characters (players) within the images, and perform operations corresponding to the recognized movements.

The video camera device 700 encodes the captured digital video signal at the video signal encoding unit 610 in the built-in transmission unit 600 with the encoding method according to the present invention, modulates the code stream at the digital modulation unit 611, and provides it to the wireless modulation unit 612 for transmission from antenna 613. The transmitted radio wave is received by the antenna 620 of the receiving device 601 built in the main unit 701 of the home game machine, and the received signal is supplied to the digital demodulation unit 623 via the wireless module unit 621 and the front end unit 622 . The received signal is demodulated into a code stream at the digital demodulation unit 623 and supplied to the video signal decoding unit 624 . The video signal decoding unit 624 decodes the supplied code stream with the decoding method according to the present invention, and outputs a baseband digital video signal.

The baseband digital video signal output from the video signal decoding unit 624 is sent via, for example, a bus in the main unit 701 of the home game machine, and temporarily stored in the RAM. When the digital video signal stored in the RAM is read out following a predetermined program, the CPU can detect the movement of a character in an image provided by the digital video signal, and use the image within the game.

Since the delay time from when an image is captured by the video camera device 700 and the obtained digital video signal is encoded until the code stream is decoded at the main unit 701 of the home game machine and an image is obtained there is short, the home game machine The responsiveness of the game software running on the main unit 701 to the movement of the player is improved, thereby improving the operability of the game.

Note that such a video camera device 700 used with a home game machine often has a simple configuration due to limitations in price, size, etc., and it must be assumed that a CPU and a large-capacity memory with high processing power such as a computer device cannot be implemented. .

That is to say, in general, the video camera device 700 is a peripheral device of the main unit 701 of the home game machine, which is only necessary for playing games using the video camera device 700, not on the main unit 701 of the home game machine necessary devices. In this case, the video camera device 700 is often sold separately from the main unit 701 of the home game machine (so-called option sold separately). In this case, installing a high-capacity CPU and a memory with a large storage capacity in the video camera device 700 to sell at a high price generally results in a decrease in the number of sold units. In this case, this would reduce the number of games sold using the video camera device 700, which would result in lower revenue. Also, especially for home games, ownership generally strongly affects the number of units sold, so a low ownership rate of the video camera device 700 may result in an even lower number of units sold.

On the other hand, selling a large number of video camera units 700 at a low price to improve the ownership rate can improve the number of sales of home games using the video camera unit 700 and improve their popularity, and it can further be expected to bring support for the main unit 701 of home game machines. incentives for further purchases. Also from this point of view, the video camera device 700 is often preferably a simple configuration.

In this case, an arrangement can be conceived in which wavelet transformation is performed at a low division level at the video signal encoding unit 610 of the transmission unit 600 built in the video camera device 700 . This reduces the need for memory capacity used with the coefficient rearrangement buffer unit.

Also, an arrangement can be conceived in which the configuration of the image encoding device exemplarily shown in FIG. 30 , which has been described with reference to the third embodiment, is applied to the video signal encoding unit 610 . Further, applying the configuration of the image encoding device exemplarily shown in FIG. 33 to the video signal encoding unit 610 eliminates the need to perform rearrangement processing of wavelet transform coefficient data on the video signal encoding unit 610 side, so the video camera The load on the device 700 side can be further reduced, which is desirable. In this case, it is necessary to use the image decoding device exemplarily shown in FIG. 34 as the video signal decoding unit 624 in the receiving device 601 built in the main unit 701 side of the home game machine.

Note that the video camera device 700 and the main unit 701 of the home game machine have been described above as being connected by wireless communication, but this arrangement is not limited to this example. That is, the video camera device 700 and the main unit 701 of the home game machine can be connected by a cable via an interface such as USB, IEEE 1394, or the like.

As described above, another great advantage of the encoding method (or decoding method) according to the present invention is that it is easy to process, and the load and processing time are small, so it can be applied to various forms and can be easily applied to various purposes (i.e. high versatility).

The series of processing described above can be realized by hardware or can be realized by software. In the case where the series of processing is realized by software, the program constituting the software is installed from a program recording medium in a computer in which dedicated hardware is built, or in a general-purpose computer, or an information processing system composed of a plurality of devices In the information processing device of the system, it is possible to execute various functions by various types of programs installed therein.

Fig. 40 is a block diagram showing an example of the configuration of an information processing system for executing the above-described series of processing by a program.

As shown in FIG. 40, an information processing system 800 is configured with an information processing device 801, a storage device 803 connected to the information processing device 801 through a PCI bus 802, and VTR 804-1 to VTR 804-S as a multi-tape video recorder (VRT). , and a system of a mouse 805, a keyboard 806, and an operation controller 807 for the user to perform its operation input, and is a system that executes image encoding processing, image decoding processing, and the like as described above by installed programs.

The information processing device 801 of the information processing system 800 can, for example, encode moving image content stored in a large-capacity storage device 803 configured with RAID (Redundant Array of Independent Disks), and store the obtained encoded data in the storage device In 803, the coded data stored in the storage device 803 is decoded, and the obtained decoded image data (moving image content) is stored in the storage device 803, and the coded data or the decoded image data are stored by the VTR 804 -1 to VTR 804-S to record on tape, etc. And, the information processing device 801 is arranged to ingest into the storage device 803 moving image contents recorded in video tapes mounted to the VTR 804-1 to VTR 804-S. At this time, an arrangement may be made in which the information processing device 801 encodes moving image content.

The information processing device 801 has a microprocessor 901, GPU (Graphics Processing Unit) 902, XDR (Extreme Data Rate)-RAM 903, South Bridge 904, HDD 905, USB interface (USB I/F) 906, and sound input/output Codec 907.

The GPU 902 is connected to the microprocessor 901 via a dedicated bus 911. XDR-RAM 903 is connected to microprocessor 901 via dedicated bus 912. South bridge 904 is connected to I/O controller 944 of microprocessor 901 via a dedicated bus. HDD 905, USB interface 906, and sound input/output codec 907 are also connected to South Bridge 904. A speaker 921 is connected to the sound input/output codec 907 . Also, a display 922 is connected to the GPU 902.

A mouse 805, a keyboard 806, VTRs 804-1 to 804-S, a storage device 803, and an operation controller 807 are also connected to the South Bridge 904 via the PCI bus 802.

The mouse 805 and the keyboard 806 receive operation input from the user, and supply a signal indicating the content of the operation input from the user to the microprocessor 901 via the PCI bus 802 and the south bridge 904 . The storage device 803 and the VTRs 804-1 to 804-S can record and play back predetermined data.

A drive 808 on which a removable medium 811 such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory is appropriately mounted is further connected to the PCI bus 802, and the computer program read therefrom It is installed in HDD 905 as needed.

The microprocessor 901 is a multi-core configuration in which a general-purpose main CPU core 941 for executing a basic program such as an OS (Operating System), as a plurality of ( In this case, eight) sub CPU cores 942-1 to 942-8 of RISC (Reduced Instruction Set Computer) type signal processing processors for executing, for example, The storage controller 943 for storage control of the XDR-RAM 903, and the I/O (input/output) controller 944 for managing the input and output of data of the south bridge 904 are integrated on a single chip, for example, to realize a 4 [GHz] operating frequency.

At startup, the microprocessor 901 reads out the necessary application program stored in the HDD 905 based on the control program stored in the HDD 905, and expands it into the XDR-RAM 903, and based on the application program and the operator's The operation executes necessary control processing sequentially.

And, by executing the software, the microprocessor 901 can, for example, realize the image encoding process and the image decoding process of the above-mentioned embodiments, provide the code stream obtained as decoding to the HDD 905 via the south bridge 904 for storage, and execute the code stream obtained as the decoding result. The transmission of the playback picture data of the obtained moving image content to the GPU 902 for display on the display 922, and the like.

Although how to use the CPU cores of the microprocessor 901 is optional, an arrangement may be made in which, for example, the main CPU core 941 performs processing related to control of image encoding processing and image decoding processing, and eight sub CPU cores 942 -1 to the sub CPU core 942-8, for example, simultaneously and in parallel as described with reference to FIG. ,etc. At this time, an arrangement in which the main CPU core 941 distributes processing to eight sub CPU cores 942-1 to sub CPU cores in increments of row blocks (partitions) is carried out in the same manner as the case described with reference to FIG. Each of the CPU cores 942-8 thereby simultaneously executes image encoding processing and image decoding processing in parallel in increments of line blocks. That is, the efficiency of image encoding processing and image decoding processing can be improved, the delay time of overall processing can be reduced, and the load necessary for processing, processing time, and memory capacity can be reduced. Of course, each processing can also be performed by other methods.

For example, an arrangement may be made in which part of the eight sub-CPU cores 942-1 to 942-8 of the microprocessor 901 performs encoding processing and the remaining part performs decoding processing simultaneously and in parallel.

And, for example, in the case where an independent encoder or decoder, or a codec processing device is connected to the PCI bus 802, the eight sub-CPU cores 942-1 to 942-8 of the microprocessor 901 can be connected via the south bridge 904 and PCI bus 802 control the processing performed by these devices. Furthermore, in the case where a plurality of such devices are connected, or in the case where the devices include a plurality of decoders or encoders, the eight sub CPU cores 942-1 to 942-8 of the microprocessor 901 can Control is performed such that multiple decoders or encoders share the process.

At this time, the main CPU core 941 manages the operations of the eight sub-CPU cores 942-1 to 942-8, assigns processing to each CPU core, retrieves processing results, and so on. Furthermore, the main CPU core 941 also executes processing other than those executed by these sub CPU cores. For example, the main CPU core 941 accepts commands supplied from the mouse 805, the keyboard 806, or the operation controller 807 via the south bridge 904, and executes various types of processing corresponding to the commands.

The GPU 902 performs final unwrapping processing on pasting textures and thus play pictures for playing moving picture content for display on the display 922, and also hosts multiple play pictures for performing when displaying moving picture content on the display 922 at a time The load on the microprocessor 901 can be relieved by a function of coordinate transformation calculation for still images of still image contents, processing for enlarging/reducing playback pictures of moving image contents and still images of still image contents, and the like.

The GPU 902, under the control of the microprocessor 901, performs predetermined signal processing on the supplied picture data of moving picture content and still image of still picture content, and transmits the picture data and image data obtained as a result thereof to the display 922 , and the image signal is displayed on the display 922.

Now, the playback pictures of a plurality of moving image contents decoded in parallel by the eight sub-CPU cores 942-1 to 942-8 of the microprocessor 901 are transmitted to the GPU 902 via the bus 911. The transmission speed is, for example, Up to 30 [Gbyte/sec], making it possible to smoothly display even complex picture images with special effects at high speed.

On the other hand, the microprocessor 901 subjects the audio data of the picture data and audio data of the moving image content to audio mixing processing, and passes the edited audio data obtained as a result thereof via the south bridge 904 and the sound input/output code 907 to the speaker 921 to output audio based on the audio data from the speaker 921.

In the case of realizing the above-described series of processes by software, programs constituting the software are installed from a network or a recording medium.

This recording medium includes not only the removable medium 811 shown in FIG. Optical discs (including MDs), semiconductor memories, etc., but also HDD 905 storing programs to be distributed to users in a state of being mounted to the apparatus main unit, storage device 803, etc. Of course, the storage medium may also be a semiconductor memory such as ROM or flash memory.

In the above, it has been described that the microprocessor 901 is configured with eight sub-CPU cores, but the present invention is not limited thereto, and the number of sub-CPU cores is optional. Also, for the microprocessor 901, an arrangement may be made in which a CPU configured with a single core (one core) instead of a main CPU core and a sub CPU core is used. Also, instead of the microprocessor 901, a plurality of CPUs may be used, and a plurality of information processing devices may be used (ie, programs for executing the processing of the present invention are jointly executed under the operation of a plurality of devices).

Although the steps describing the program stored in the program recording medium in this specification can of course be executed in the described time series, they are not limited to this time series and may be executed in parallel or individually.

In addition, the system used in this specification refers to the entirety of equipment configured with a plurality of devices (devices).

Note that, in the above, the configuration described as a single device may be divided so as to be configured with a plurality of devices. On the contrary, in the above, the configuration described as a plurality of devices may be combined so as to be configured with a single device. Also, the configuration of each device may have configurations other than those described above added thereto. Furthermore, a part of the configuration of one device may be included in the configuration of another device as long as the configuration and operation of the overall system are substantially the same.

Claims (18)

1. An encoding device, which is used to encode coefficient data, said encoding device comprising: a maximum significant digit generating section for taking, as the maximum significant digit of the coefficient data, the significant digit of a numerical value whose absolute value is the largest among numerical values represented by each of the coefficient data, and generating an index indicating the maximum Code of significant digits; an absolute value generating section for generating a code indicating an absolute value of a numerical value represented by the coefficient data; and a code generation section for generating a code indicating a sign of a numerical value represented by the coefficient data, Wherein, as the code indicating the maximum significant digit number, the maximum significant digit number generating section generates a code indicating whether the maximum significant digit number has been changed, and when the maximum significant digit number is changed, the maximum significant digit number The number of digits generation section generates a code indicating whether the maximum significant number of digits has increased or decreased and a code indicating the amount of change of the maximum significant number of digits. 2. The encoding device according to claim 1, further comprising a code linking section for linking the code indicating the maximum significant digit generated by the maximum significant digit generating section, the code having been generated by the absolute value generating section The generated code indicating the absolute value and the code indicating the sign generated by the code generating unit generate coded data. 3. The encoding device according to claim 1, wherein the maximum significant digit generation section repeats the process of generating a code indicating the maximum significant digit number of the coefficient data arranged consecutively, and generating a code indicating the number of significant digits from the coefficient data The code of the maximum significant number of digits of the coefficient data starting from the further continuous arrangement of the coefficient data. 4. The encoding device according to claim 1, wherein the absolute value generation section generates a code indicating a value from the lowest-order digit to the largest significant digit of the numerical value represented by the coefficient data as an indication of the value represented by Code for the absolute value of the numerical value represented by the coefficient data. 5. The encoding device according to claim 1, wherein the absolute value generating section generates codes indicating absolute values of numerical values represented by the coefficient data in parallel. 6. The encoding device according to claim 1 , further comprising a zero code generating section for generating a code indicating whether all numerical values represented by the coefficient data are 0, Wherein, if the value indicated in the code indicating whether it is 0 or not generated by the zero code generation section indicates that not all values represented by the coefficient data are 0, the coefficient data is encoded. 7. The encoding device of claim 1, further comprising: a wavelet transform section for performing wavelet transform on image data to generate the coefficient data; a storage section for storing coefficient data generated by the wavelet transformation section; and a coefficient rearranging section for rearranging in advance the coefficient data stored in the storage section in an order in which inverse wavelet transformation is to be performed on the coefficient data. 8. The encoding device of claim 1, further comprising: a filter section for performing filter processing on the image data of each line block of at least the lowest frequency band component subband, wherein the line block includes generating an equivalent line The coefficient data necessary for several lines of image data; a storage section for storing coefficient data generated by the filter section for each line block; a coefficient rearranging section for rearranging in advance the coefficient data stored by the storage section in an order in which synthesis processing is to be performed for each line block, wherein on the decoding side, coefficients of a plurality of subbands divided by frequency band The data are to be synthesized to generate image data. 9. The encoding device according to claim 8, wherein the coefficient rearranging section rearranges the coefficient data in an order from a low-band component to a high-band component for each line block. 10. An encoding method, which is used to encode coefficient data, said encoding method comprising the steps of: taking, as a maximum significant digit of said coefficient data, a significant digit of a numerical value whose absolute value is the largest among numerical values represented by each of said coefficient data, and generating a code indicating said maximum significant digit; generating a code indicating the absolute value of the numerical value represented by the coefficient data; and generating a code indicating the sign of the numerical value represented by said coefficient data, Wherein, as the code indicating the maximum significant digit number, a code indicating whether the maximum significant digit number has been changed is generated, and in a case where the maximum significant digit number is changed, a code indicating whether the maximum significant digit number has been increased or decreased is generated code and a code indicating the amount of change in the maximum significant digit. 11. A decoding device for decoding encoded data and generating coefficient data, the decoding device comprising: a maximum significant digit calculation section for decoding a code indicating a maximum significant digit number, wherein the maximum significant digit number of the coefficient data is determined by each of the coefficient data The number of significant digits of the value with the largest absolute value among the represented values; an absolute value calculation section for decoding a code indicating an absolute value of a numerical value represented by the coefficient data, and calculating an absolute value of the numerical value represented by the coefficient data; a code generation section for decoding a code indicating a sign of a numerical value represented by the coefficient data, and generating data of a sign of a numerical value represented by the coefficient data; and a coefficient data generation section for indicating a sign based on the maximum significant digit calculated by the maximum significant digit calculation section, the absolute value calculated by the absolute value calculation section, and the code generation section , decode the encoded data and generate coefficient data, As a code indicating the maximum significant digit number, the maximum significant digit calculation section decodes a code indicating whether the maximum significant digit number has been changed, and in a case where the maximum significant digit number has been changed, the maximum significant digit calculation section The section decodes a code indicating whether the maximum significant digit has increased or decreased, and a code indicating a change amount of the maximum significant digit, and calculates the maximum significant digit based on a decoding result. 12. The decoding apparatus according to claim 11 , further comprising a code separation section for separating said encoded data into a code indicating said maximum significant number of digits, a code indicating said absolute value, and a code indicating said The code for the above symbol. 13. The decoding device according to claim 11 , wherein the maximum significant digit calculation section repeats the process of decoding codes indicating the maximum significant digits of the coefficient data that are continuously arranged, and decoding codes indicating the maximum significant digits from the coefficient data. The code of the maximum significant number of digits of the coefficient data starting from the further continuous arrangement of the coefficient data. 14. The decoding device according to claim 11, wherein the absolute value calculation section decodes codes indicating absolute values of numerical values represented by the coefficient data in parallel. 15. The decoding apparatus according to claim 11 , further comprising: a decoding section for decoding the encoded data, and generating the coefficient data obtained by performing wavelet transformation; a storage section for storing the coefficient data generated by the decoding section; and a coefficient rearranging section for rearranging the coefficient data stored by the storage section in an order in which inverse wavelet transformation is to be performed on the coefficient data. 16. The decoding apparatus of claim 11 , further comprising: a decoding section for decoding the encoded data, and generating the coefficient data included in a plurality of subbands; a storage section for storing, for each line block of at least the lowest frequency band component subband, the coefficient data generated by the decoding section, wherein the line block includes a number of line images necessary for generating coefficient data equivalent to one line data; and a coefficient rearranging section for rearranging, for each line block, the coefficient data stored by the storage section in an order in which synthesis of coefficient data of a plurality of subbands divided by frequency band is to be performed. 17. A decoding method for decoding encoded data and generating coefficient data, said decoding method comprising the steps of: decoding a code indicating a maximum number of significant digits in said encoded data, and calculating said maximum significant number of digits, wherein a maximum significant number of said coefficient data is among numerical values represented by each of said coefficient data The number of significant digits of the value with the largest absolute value; decoding a code indicating an absolute value of a numerical value represented by the coefficient data, and calculating an absolute value of the numerical value represented by the coefficient data; decoding a code indicating a sign of a numerical value represented by the coefficient data, and generating data indicating a sign of a numerical value represented by the coefficient data; and decoding the encoded data and generating coefficient data based on the calculated maximum significant digit number, the calculated absolute value, and the generated data indicating a sign, As the code indicating the maximum significant number of digits, decoding a code indicating whether the maximum significant digit has been changed, and if the maximum significant digit has been changed, decoding a code indicating whether the maximum significant digit has been increased or decreased, and A code indicating a change amount of the maximum significant digit, and the maximum significant digit is calculated based on a decoding result. 18. A transmission system comprising: encoding means for encoding coefficient data; and decoding means for decoding encoded data obtained by encoding the coefficient data; wherein said encoded data is transmitted between said encoding means and said decoding means; The encoding device includes: a maximum significant digit generating section for taking, as the maximum significant digit of the coefficient data, the significant digit of a numerical value having the largest absolute value among the numerical values represented by each of the coefficient data, and generating an index indicating the Code with the largest number of significant digits; an absolute value generating section for generating a code indicating an absolute value of a numerical value represented by the coefficient data; and a code generation section for generating a code indicating a sign of a numerical value represented by the coefficient data; a code linking section for linking the code indicating the maximum significant digit which has been generated by the maximum significant digit generating section, the code indicating an absolute value which has been generated by the absolute value generating section, and the code generated by the code The code indicating the symbol generated by the section, thereby generating the encoded data, and a transmission section for transmitting the encoded data generated by the code linking section; Wherein, as the code indicating the maximum significant digit number, the maximum significant digit number generating section generates a code indicating whether the maximum significant digit number has been changed, and when the maximum significant digit number is changed, the maximum significant digit number the number of digits generation section generates a code indicating whether the maximum significant number of digits has increased or decreased and a code indicating a change amount of the maximum significant number of digits; And the decoding device includes: a receiving unit for receiving the encoded data transmitted by the transmitting unit; a code separation section for separating the encoded data received by the reception section into a code indicating the maximum significant number of digits, a code indicating the absolute value, and a code indicating the sign, a maximum significant digit calculation section for decoding the code indicating the maximum significant digit number separated by the code separation section, and calculating the maximum significant digit number; an absolute value calculation section for decoding the code indicating the absolute value separated by the code separation section, and calculating an absolute value of a numerical value represented by the coefficient data; a code generating section for decoding the code indicating the symbol separated by the code separating section, and generating data indicating the symbol represented by the coefficient data; and a coefficient data generating section for indicating a sign based on the maximum significant digit calculated by the maximum significant digit calculating section, the absolute value calculated by the absolute value calculating section, and the code generating section , decode the encoded data and generate coefficient data, As a code indicating the maximum significant digit number, the maximum significant digit calculation section decodes a code indicating whether the maximum significant digit number has been changed, and in a case where the maximum significant digit number has been changed, the maximum significant digit calculation section The section decodes a code indicating whether the maximum significant digit has increased or decreased, and a code indicating a change amount of the maximum significant digit, and calculates the maximum significant digit based on a decoding result.

Images